Download as pdf or txt
Download as pdf or txt
You are on page 1of 184

Thermodynamic Modeling of Aqueous Electrolytes and Ionic Liquids (ILs) with

Electrolyte Non Random Two Liquid (eNRTL) Model

by

Nazir Hossain, B.Sc, M.Sc

A Dissertation

In

Chemical Engineering

Submitted to the Graduate Faculty


of Texas Tech University in
Partial Fulfilment of
the Requirements for
the Degree of

DOCTOR OF PHILOSOPHY

Approved

Dr. Chau-Chyun Chen


Chair of Committee

Dr. Fazle Hussain

Dr. Rajesh Khare

Dr. Nurxat Nuraje

Mark Sheridan
Dean of the Graduate School

August 2018
Copyright 2018, Nazir Hossain
Dedication:
I would like to dedicate this thesis to my parents who are my constant source of
inspiration to do hard work.
Texas Tech University, Nazir Hossain, August 2018

ACKNOWLEDGEMENTS
At first I would like to show my gratitude to the Almighty who is the creator of

the universe and all the lives here. He gives me the knowledge and strength to accomplish

this dissertation.

Now I would like to take a moment and thank everyone who gave me the

necessary supports in this long journey. I would like to thank my Ph.D. supervisor, Dr.

Chau-Chyun Chen, who has been a great mentor. I am so lucky that I have such a

wonderful adviser who is always available to give his time and advice. He teaches me to

be an independent thinker. I would like to thank my all of committee members, Dr. Fazle

Hussain, Dr. Rajesh Khare, and Dr. Dr. Nurxat Nuraje, for their valuable suggestions

during my qualifying examination. I would like to thank all of the faculty members in the

Chemical Engineering Department.

I would like to thank Dr. Khare group for the collaboration with our group. I

learned molecular dynamics simulation from the course taken by Dr. Khare. He is a great

teacher. I would also appreciate the help of Aswhin Ravichandran and Rafikul Islam to

conduct molecular dynamics simulation study in support of my thermodynamic works. I

had opportunity of working with Dr. Sanjoy Bhattacharia, Sheik Tanveer, and Yue Yu. I

would like to thank them for their valuable inputs in my research. I would also thank Hla

Tun, Harnoor, Pradeep, Sina, Soraya, Nguyen, Coni, Meng, Yuan, Yifan, Islam, Toni,

Ben, Rex, Rajisi, and Michael for being amazing lab mates who always support and

inspire me for the last 4 years. I am really lucky to have a wonderful research group. I

would like to thank the J.F Maddox Foundation for the financial support for this research

work.
ii
Texas Tech University, Nazir Hossain, August 2018

I thank my family including my parents, siblings, parents-in-law, and brother-in-

law for their constant support and prayers. I am thankful to my wife, Nowshin Tabassum

Rahman, for her prayers and mental support to continue my research under stress.

Finally, I am indebted to the Bangladeshi Lubbock community for being my family away

from my home country. I want to thank each and every person of this community for

their all sorts of support on the way of accomplishing one of the most significant

achievements of my life.

iii
Texas Tech University, Nazir Hossain, August 2018

TABLE OF CONTENTS

ACKNOWLEDGEMENTS …...………………………………………………………….ii
ABSTRACT ……………………………………………………………………………..vii

LIST OF TABLES ……………………………………………………………………… ix

LIST OF FIGURES ………………………………………………………………………xi

CHAPTER 1 INTRODUCTION .........................................................................................1


1.1 Background .......................................................................................................... 1

1.2 Content of the Thesis............................................................................................ 7

CHAPTER 2 REVISITING ELECTROLYTE THERMODYNAMIC MODELS:


INSIGHTS FROM MOLECULAR SIMULATIONS .......................................................15
2.1 Abstract .............................................................................................................. 15

2.2 Introduction ........................................................................................................ 16

2.3 Theoretical and Computational Details .............................................................. 19

2.3.1 Theoretical Background .............................................................................. 19

2.3.2 Simulation Details....................................................................................... 21

2.3.3 Thermodynamic Modeling.......................................................................... 23

2.4 Results and Discussion ....................................................................................... 23

2.4.1 Mean Ionic Activity Coefficient Predicted by the eNRTL Model

Qualitatively Agrees with MD Calculations and Experimental Data........................ 23

2.5 Discussion and Conclusions ............................................................................... 28

CHAPTER 3 TEMPERATURE DEPENDENCE OF INTERACTION PARAMETERS


IN ELECTROLYTE NRTL MODEL ...............................................................................44
3.1 Abstract .............................................................................................................. 44
iv
Texas Tech University, Nazir Hossain, August 2018

3.2 Introduction ........................................................................................................ 45

3.3 Thermodynamic Framework .............................................................................. 49

3.4 Temperature Dependence of Interaction Parameters ......................................... 55

3.5 Conclusions ........................................................................................................ 61

CHAPTER 4 THERMODYNAMIC MODELING OF AQUEOUS LI+–NA+–K+–


MG2+−SO42− QUNIARY SYSTEMS ................................................................................86
4.1 Abstract .............................................................................................................. 86

4.2 Introduction ........................................................................................................ 87

4.3 Thermodynamic Frameworks ............................................................................ 90

4.4 Li2SO4 + H2O Binary ......................................................................................... 91

4.5 Ternary Systems ................................................................................................. 94

4.5.1 Li2SO4 + Na2SO4 + H2O Ternary ............................................................... 94

4.5.2 Li2SO4 + K2SO4 + H2O Ternary ................................................................. 95

4.5.3 Li2SO4 + MgSO4 + H2O Ternary ................................................................ 96

4.6 Quaternary System ............................................................................................. 97

4.6.1 Li2SO4 + K2SO4 + MgSO4 + H2O Quaternary............................................ 97

4.7 Conclusions ........................................................................................................ 97

CHAPTER 5 DISSOCIATION BEHAVIOR OF IONIC LIQUIDS IN WATER:


THERMODYNAMIC MODELING AND MOLECULAR DYNAMICS
SIMULATIONS ..............................................................................................................118
5.1 Abstract ............................................................................................................ 118

5.2 Introduction ...................................................................................................... 119


v
Texas Tech University, Nazir Hossain, August 2018

5.3 Experimental Data ............................................................................................ 121

5.4 Symmetric eNRTL Model with Conventional Dissociation Chemistry .......... 121

5.5 Molecular Dynamics Simulation...................................................................... 125

5.5.1 Prior Works ............................................................................................... 125

5.5.2 Simulation Methods .................................................................................. 125

5.5.3 Structural Properties.................................................................................. 127

5.5.4 Clustering of Ionic Liquids ....................................................................... 128

5.6 Symmetric eNRTL Model with New Dissociation Chemistry ........................ 129

5.7 Conclusions ...................................................................................................... 132

CHAPTER 6 CONCLUSIONS AND FUTURE WORKS ..............................................165

vi
Texas Tech University, Nazir Hossain, August 2018

ABSTRACT

Accurate and consistent thermodynamic model is essential to address the non-

ideality of the aqueous electrolyte and ionic liquids (ILs). The necessity of

thermodynamic modeling for aqueous electrolytes arises in the mass and energy balance

in the industrial applications such as desalination of highly saline water, treatment of

produced water from the hydraulic fracturing, membrane separation, solubility

predictions, salt extraction from the salt-lake brine, etc. The most widely used

thermodynamic models are the Pitzer model and the electrolyte nonrandom two liquid

(eNRTL) model. The interaction parameters of the Pitzer model are nonlinear functions

of concentration and temperature. On the other hand, the eNRTL model parameters are

only a well-defined function of temperature. In this study, we first assess the

concentration dependence of the interaction parameters of both Pitzer and eNRTL model

for NaCl + H2O water system. The Pitzer model and the eNRTL model show a deviation

in the extrapolation of the mean ionic activity coefficient beyond 6 molal concentration.

We calculated the mean ionic activity coefficient in a wide range of concentration based

on molecular simulations and Kirkwood-Buff theory. The eNRTL model shows the

correct asymptotic behavior of the mean ionic activity coefficient which is supported by

molecular dynamics simulations and supersaturation experimental data, while the Pitzer
model diverges very quickly beyond the salt saturation. Then we expressed the

interaction parameters of the eNRTL model as a Gibbs-Helmholtz type equation which

has less adjustable parameters than the Pitzer model. We successfully applied the eNRTL

model for aqueous Li+–Na+–K+–Mg2+−SO42− quinary system to correlate and predict the

thermodynamic properties accurately from 273.15-573.15 K and concentration up to

saturation.

vii
Texas Tech University, Nazir Hossain, August 2018

The applicability of the eNRTL model is extended to the aqueous ionic liquids

(ILs). Ionic liquid has gained a lot of attention in the past few decades for their negligible

volatility, which would make them “green solvent”. The dissociation behavior of the

ionic liquids in water is a particular interest in this study. The eNRTL model and

COSMO-SAC model fail to satisfy the experimental dissociation extent data of ionic

liquids in water with conventional partial dissociation chemistry. We propose a new

dissociation chemistry based on the finding from the molecular dynamics simulation. The

eNRTL model, with new dissociation chemistry, correlates the dissociation extent and
vapor-liquid equilibrium (VLE) data throughout the concentration range.

viii
Texas Tech University, Nazir Hossain, August 2018

LIST OF TABLES

2.1 Kirkwood-Buff derived force field parameters for NaCl [30] and
SPC/E parameters for water [41]. ......................................................................... 30
2.2 Number of NaCl and water molecules at different concentrations used
in molecular dynamics simulations....................................................................... 31
2.3 The values of activity coefficient of salt and water on different scales
predicted by MD simulations. The superscripts 𝑀, 𝑚, and 𝑥
denote molarity, molality, and mole fraction scales, respectively.
The water activity coefficient is based on the apparent mole
fraction of water. ................................................................................................... 32
3.1 Regressed Gibbs energy and enthalpy terms of interaction parameters
from mean ionic activity coefficient [17] and excess enthalpy [19]
of electrolytes at 298.15 K with 𝛼 = 0.2 .............................................................. 62
3.2 Thermodynamic constants for water and ions at 298.15 K and 0.1 MPa
[19] ........................................................................................................................ 66
3.3 Regressed heat capacity terms of interaction parameters from heat
capacity data [21-26] at 298.15 K with 𝛼 = 0.2 ................................................... 67
4.1 Experimental thermodynamic property data for Li2SO4 + H2O binary. ..................... 99
4.2 eNRTL binary interaction parameter, τij, for aqueous Li+-Na+-K+-
Mg2+-SO42- quinary system. ................................................................................ 100
4.3 Thermodynamic constants for salts at 298.15 K and 0.1 MPa. ................................ 102
4.4 Thermodynamic constants for water and ions at 298.15 K and 0.1
MPaa. ................................................................................................................... 103
4.5 Experimental solubility data for the Li2SO4 in aqueous solution in
presence of other salts. The bullets show what salts are present in
the solution. ......................................................................................................... 104
5.1 Literature data and corresponding MRD (%). .......................................................... 133
5.2 Binary interaction parameters of eNRTL model with chemistry I for
[emim][EtSO4], [emim][TFA], and [emim][TfO] in water. ............................... 135
5.3 Equilibrium constants of Ionic Liquids (ILs) for chemistry I. .................................. 136
5.4 The simulation box composition of the binary mixtures of water and
ionic liquids......................................................................................................... 137
5.5 The simulated and experimental density (g/cc) of the pure ionic liquids
at 298.15 K and atmospheric pressure. ............................................................... 138
5.6 The number of cluster elements in clusters size up to 5. .......................................... 139

ix
Texas Tech University, Nazir Hossain, August 2018

5.7 Dissociation extent (ξ) from molecular dynamics simulation of 0.5


mol% ILs in water up to cluster size 3. ............................................................... 140
5.8 Binary interaction parameters of eNRTL model with Chemistry II for
[emim][EtSO4], [emim][TFA], and [emim][TfO] in water. ............................... 141
5.9 Equilibrium constants of Ionic Liquids (ILs) for chemistry II ................................. 142

x
Texas Tech University, Nazir Hossain, August 2018

LIST OF FIGURES

2.1 Procedure for obtaining the value of Kirkwood-Buff integral is


illustrated at 12 molal NaCl concentration. The dashed black line
indicates the region up to which the linear fit was performed to
calculateGij∞. ....................................................................................................... 34
2.2 Mean ionic activity coefficient of NaCl as a function of salt
concentration at 298.15 K and 1 bar on mole fraction scale. The
predictions from Pitzer [17, 18] and eNRTL [5] models are
compared with that calculated from KB theory and experimental
data (Robinson and Stokes [34] and Tang et al. [36]). 𝛾 ±
calculations of Weerasinghe and Smith [30] are also shown. .............................. 35
2.3 Mean ionic activity coefficient of NaCl as a function of salt
concentration at 298.15 K and 1 bar on molality scale. The
predictions from Pitzer [17, 18] and eNRTL [5] models are
compared with that calculated from KB theory and experimental
data (Robinson and Stokes [34] and Tang et al. [36]). 𝛾 ±
calculations of Weerasinghe and Smith [30] are also shown. .............................. 36
2.4 Local and the electrostatic contribution to the mean ionic activity
coefficient of NaCl over a wide range of salt concentrations as
predicted by the eNRTL model. For comparison, the activity
coefficient predicted by the Pitzer model is also shown. ...................................... 37
2.5 Comparison of water activity coefficient as a function of salt
concentration at 298.15 K and 1 bar as predicted by the Pitzer and
the eNRTL models. Activity coefficient values of water predicted
using KB theory are also shown in red symbols and the
experimental measurements of Tang et al. [36] and Chan et al. [37]
are shown in black symbols. ................................................................................. 38
2.6 Comparison of water activity as a function of salt concentration at
298.15 K and 1 bar as predicted by the Pitzer and the eNRTL
models. The activity values of water predicted using KB theory are
also shown in red symbols and the experimental measurements of
Tang et al.[36] and Chan et al. [37] are shown in black symbols. ........................ 39
3.1 Temperature dependence of interaction parameters of Pitzer model for
NaCl + H2O and Na2SO4 + H2O binary systems; (a) NaCl + H2O: (
) 10*β(0), ( ) β(1), ( ) 100*Cγ; (b) Na2SO4 + H2O: (
) 10*β , (
(0)
)β ,(
(1)
) 100*Cγ .......................................................... 68
3.2 Temperature dependence of eNRTL binary interaction parameters for
NaCl + H2O and Na2SO4 + H2O binary systems [16]; ( ) H2O -
(Na+ Cl-), ( ) (Na+ Cl-) - H2O, ( ) H2O - (Na+ SO42-), (
) (Na+ SO42-) - H2O ..................................................................................... 69

xi
Texas Tech University, Nazir Hossain, August 2018

3.3 ExperimentalChart Title


mole fraction scale mean ionic activity coefficients of 1-
Chart Title
1 electrolytes at 298.15 K [17] .............................................................................. 70
2 Chart Title
Chart Title
3.4 Mean 1.2
Chart Title ionic
1.2
activity coefficient of 1-1 electrolytes at 298.15 K
1.2 calculated by the PDH term and the lc term with various values for
1 1 ca,
(∆gw, 1 ∆gca, w): ( ) PDH; ( ) (6, −3); ( ) (8,
1 −4); ( ) (8.845,Chart −4.534) (NaCl));
Title ( ) (10, −6);( )
(12, −6)
Chart Title
0.8.................................................................................................................. 71
8 Chart Title 0.8
Chart Title Chart Title
1.2 0.8 1.2Gibbs energy of aqueous 1-1 electrolyte solution at
3.5 Symmetric0.6excess
1.2 298.15
0.6 K calculated by the PDH term and the lc term with varying
.2
6
1 0.6 values0.4 1 ca, ∆gca, w): (
for (∆gw, ) PDH; ( ) (6, −3) (
1 1 ) (6.9315, −3.4657); ( ) (8, −4); ( ) (8.845,
4 0.4 0.2 (NaCl)); (
−4.534) ) (10, −5) ....................................................................... 72
0.8 0.4 0.8
.8 0.8 0
3.6 eNRTL model results for symmetric excess Gibbs energy of some
2 0.2 0.2 0
aqueous electrolytes
0.2 0.4 0.6 0.8 1 1.2
at 298.15 K .......................................................................... 73
0.6
0.4 0.6 0.8 0.6 1 1.2
.6
0.6 3.7 Experimental excess enthalpy of some 1-1 electrolytes in unsymmetric
0 0 0
reference state
.4 0 0.4 0.4at 298.15 K [19] ............................................................................ 74
0.2 0 0.4 0 0.2 0.6 0.2
0.4 0.8 0.4 0.6 1 0.80.6 1.2 10.8 1.2 1 1.2
0.4 3.8 Pitzer-Debye-Hückel prediction of excess enthalpy of aqueous
.2 0.2 electrolytes at0.2
298.15 K: ( ) 1-1 electrolytes, ( ) 1-2/2-1, (
) 2-2, ( ) 1-3/3-1 ................................................................................. 75
0.2
0 3.9 eNRTL model prediction of excess enthalpy of aqueous NaCl solution
0 0
0 0.2 0.4at 298.15 K 0.6along with 0.8
experimental 1data [19] with
1.2 fixed values
0 0 0.2 0.4 0 0.6 0.2 0.8 0.4 1 0.6 1.2 0.8 1
.4 0.6 for Gibbs energy terms (∆gw, ca= 8.845 and ∆gca, w= −4.534)
0 0.2 0.8 0.4 1 0.6 1.2 0.8 1 1.2
and varying values for enthalpy terms (∆hw, ca , ∆hca, w) of
interaction parameters: ( ) PDH; ( ) (100, 0); ( )
(200, 0); ( ) (300, 0); ( ) (400, 0) ; ( ) (500, 0) .............................. 76
3.10 eNRTL model prediction of excess enthalpy of aqueous NaCl
solution at 298.15 K along with experimental data [19] with fixed
values for Gibbs energy terms (∆gw, ca= 8.845 and ∆gca, w=
−4.534) and varying values for enthalpy terms (∆hw, ca, ∆hca, w)
of interaction parameters: ( ) PDH; ( ) (500, −200); (
) (500, −150); ( ) (500, −100); ( ) (500, −50) ; (
) (500, 0) ...................................................................................................... 77
3.11 Experimental heat capacity data of some 1-1, 1-2, and 2-1
electrolytes at 298.15 K [21-26] ........................................................................... 78
3.12 Experimental excess heat capacity data at 298.15 K: ( ) NaCl
[23], ( ) KCl [26], ( ) Na2SO4, [25] ( ) K2SO4, (
) CaCl2 [23] and ( ) MgCl2 [23] ........................................................ 79
3.13 eNRTL model prediction of excess heat capacity of aqueous NaCl
solution at 298.15 K along with experimental data (Δ) [21] (○) [23]
with fixed values for Gibbs energy (∆gw, ca= 8.845 and ∆gca, w=
xii
Texas Tech University, Nazir Hossain, August 2018

−4.534) and enthalpy terms (∆hw, ca= 371.2and ∆hca, w= −37.3)


and changing values for heat capacity terms (∆𝑐𝑝𝑤, 𝑐𝑎, ∆𝑐𝑝𝑐𝑎, 𝑤)
of interaction parameters: ( ) (−3, 3); ( ) (−3, 2); ( )
(−3, 0) ( ) (−3, −2); ( ) (−3, −3) ............................................................. 80
3.14 eNRTL model prediction of excess heat capacity of aqueous NaCl
solution at 298.15 K along with experimental data (Δ) [21] (○) [23]
with fixed values for Gibbs energy (∆gw, ca= 8.845 and ∆gca, w=
−4.534) and enthalpy terms (∆hw, ca= 371.2and ∆hca, w= −37.3)
and changing values for heat capacity terms (∆𝑐𝑝𝑤, 𝑐𝑎, ∆𝑐𝑝𝑐𝑎, 𝑤)
of interaction parameters: ( ) (−3, 3); ( ) (−2, 3); ( )
(0, 3); ( ) (2, 3); ( ) (3, 3) ....................................................................... 81
3.15 Temperature dependence of eNRTL binary interaction parameters for
Chart Title NaCl + H2O and Na2SO4 + Chart Title systems; Comprehensive
H2O binary
Chart
1.2 Title
study [16] ( ) H2O - (Na+ Cl-), ( ) (Na+ Cl-) - H2O, (
) H2O - (Na+ SO42-), ( ) (Na+ SO42-) - H2O; This study:
+ -
1 ( ) H2O - (Na Cl ), ( ) (Na+ Cl-) - H2O, ( ) H2O -
+ 2- + 2-
(Na SO4 ), ( ) (Na SO4 ) - H2O ............................................................... 82
0.8
4.1 Osmotic coefficient of aqueous Li2SO4 solution: The solid lines show
the model results compared to the experimental data: at 298.15 K
0.6 [15] shown in symbols: (□) 298.15 K, (○) 323.15 K, ( ) 298.15
K,( ) 323.15 K, ( ) 348.15 K, ( ) 373.15 K, ( )
0.4 423.15 K.............................................................................................................. 105
4.2
0.2 Molality scale mean ionic activity coefficient of Li2SO4 in aqueous
Li2SO4 solution: The lines show the model results ( ) 298.15 K,
0.4 0.6 0.8 1 1.2
0 ( ) 323.15 K, ( ) 348.15 K, ( ) 373.15 K, ( ) 423.15
0 K. Experimental
0.2 data0.4[15] are available
0.6 at 298.15
0.8 K, and 323.15
1 K 1.2
0.2 0.4 0.6 0.8 1
which are shown in symbols (□)1.2and (○), respectively. ...................................... 106
4.3 Excess enthalpy of aqueous Li2SO solution: The solid lines show the
model results compared to the experimental data at 298.15 K [27]
shown in symbols (○).......................................................................................... 107
4.4 Vapor Pressure of aqueous Li2SO4 solution: The solid lines show the
model results compared to the experimental data shown in symbols
[30] (×) 298.15 K, (◊) 323.15 K, (♦) 338.15 K, (○) 343.15 K, (▲)
363.15 K, (Δ) 373.15 K, (■) 423.15 K, (□) 573.15 K......................................... 108
4.5 Heat Capacity of aqueous Li2SO4 solution: The solid lines show the
model results compared to the experimental data shown in symbols
[30] (○) 298.15 K, (Δ) 323.15 K, (◊) 348.15 K (×) 363.15 K, (□)
372.15 K.............................................................................................................. 109
4.6 Solubility of Li2SO4 in water: The solid line show the model results
compared to the experimental data shown in symbols (○) [28]......................... 110

xiii
Texas Tech University, Nazir Hossain, August 2018

4.7 Solubility of aqueous Li+–Na+–SO42− ternary system: The solid lines


show the model results compared to the experimental data [28]
shown in symbols:............................................................................................... 111
4.8 Solubility of aqueous Li+–K+–SO42− ternary system: The solid lines
show the model results compared to the experimental data [31]
shown in symbols: solubility at 298.15 K: (○) Li2SO4∙H2O, (◊)
Li2SO4∙K2SO4, (×) K2SO4. ( ) and ( ) are model predictions
2−
of solubility of aqueous Li –K –SO4 ternary system at 323.15 K
+ +

and 348.15 K, respectively.................................................................................. 112


4.9 Solubility of aqueous Li+–Mg2+–SO42− ternary system: The solid lines
show the model results compared to the experimental data shown
in symbols: .......................................................................................................... 113
4.10 Jãnecke phase diagram of aqueous Li2SO4 + K2SO4 + MgSO4 + H2O
quaternary system at 273.15 K: The model results and
experimental data [20] are shown with solid lines and symbols
(○), respectively. ................................................................................................. 114
5.1 The chemical structure of (a) [emim][EtSO4], (b) [emim][TFA], and
(c) [emim][TfO]. ................................................................................................. 143
5.2 Degree of dissociation of ILs in water: comparison of experimental
data in symbols with eNRTL model results (Chemistry I) in lines
at 298.15 K: ( ), ( ) [emim][EtSO4]; ( ), ( )
[emim][TFA]; and ( ), ( ) [emim][TfO] .......................................... 144
5.3 The center of mass coordinate radial distribution function (gr) of ionic
liquids studied. (a), (b), and (c) represents 0.5 mol%
[emim][EtSO4], [emim][TFA], and [emim][TfO] in water,
respectively. The legends of the plot are given in the following
way: cation-anion (red line), anion-anion (green line), and cation-
cation (black line), respectively. ......................................................................... 145
5.4 The RDF between the most acidic hydrogen atom of the cation and the
oxygen atom of the anion: (a) [emim][EtSO4], (b) [emim][TFA],
and (c) [emim][TfO]. The red and green curves represent 0.5 mol%
ILs in water and pure ILs, respectively. .............................................................. 146
5.5 Cluster size distribution of ILs containing 99.5 mol% water: (a)
[emim][EtSO4], (b) [emim][TFA], and (c) [emim][TfO]. The inset
of Figure 5.5(a) shows the cluster size distribution at the low range
of cluster size. ..................................................................................................... 147
5.6 Degree of dissociation of ILs from molecular dynamics simulation up
to cluster size 5 at 298.15 K ................................................................................ 148
5.7 Degree of dissociation of [emim][EtSO4] in water: comparison of
experimental data in symbols with eNRTL model results
xiv
Texas Tech University, Nazir Hossain, August 2018

(Chemistry II) in lines at different temperatures: ( ), ( )


298.15 K; ( ), ( ) 313.15 K; and ( ), ( )
333.15 K.............................................................................................................. 149
5.8 Degree of dissociation of [emim][TFA] in water: comparison of
experimental data in symbols with eNRTL model results
(Chemistry II) in lines at different temperatures: ( ), ( )
298.15 K; ( ), ( ) 313.15 K; and ( ), ( )
333.15 K.............................................................................................................. 150
5.9 Degree of dissociation of [emim][TfO] in water: comparison of
experimental data in symbols with eNRTL model results
(Chemistry II) in lines at different temperatures: ( ), ( )
298.15 K; ( ), ( ) 313.15 K; and ( ), ( )
333.15 K.............................................................................................................. 151
5.10 Isobaric VLE of [emim][EtSO4] + H2O binary at 0.1013 MPa:
comparison of experimental data [21] shown in symbols with
eNRTL model results (Chemistry II): ( ), ( ) eNRTL
model................................................................................................................... 152
5.11 Activity coefficient of water of [emim][EtSO4] + H2O binary at
0.1013 MPa: comparison of experimental data shown in symbols
with eNRTL model results (Chemistry II): ( ) experiment [21], (
) eNRTL model. .................................................................................... 153
5.12 VLE of [emim][EtSO4] + H2O binary at fixed mole fractions of
[emim][EtSO4]: comparison of experimental data [22] shown in
symbols with eNRTL model results (Chemistry II) shown as lines:
( ), ( ) 0.0085; ( ), ( ) 0.0709; ( ), (
) 0.1511. .............................................................................................................. 154
5.13 Isothermal VLE of [emim][TFA] + H2O binary at different
temperatures: comparison of experimental data [23] shown in
symbols with eNRTL model results (Chemistry II) shown as
dashed lines: ( ), ( ) 328.3 K; ( ), ( ) 348.2 K;
( ), ( ) 358.2 K ; ( ), ( ) 368.2 K......................................... 155
5.14 Activity coefficient of water of [emim][TFA] + H2O binary at
different temperatures: comparison of experimental data [23]
shown in symbols with eNRTL model results (Chemistry II)
shown as dashed lines: ( ), ( ) 328.3 K; ( ), (
) 348.2 K; ( ), ( ) 358.2 K ; ( ), ( ) 368.2 K. ...................... 156
5.15 Isothermal VLE of [emim][TfO] + H2O binary at different
temperatures: comparison of experimental data [23] shown in
symbols with eNRTL model results (Chemistry II) shown as dotted
lines: ( ), ( ) 323.3 K; ( ), ( ) 333.3 K; ( ), (

xv
Texas Tech University, Nazir Hossain, August 2018

) 343.3 K ; ( ), ( ) 353.3 K; ( X ), ( )
363.3 K................................................................................................................ 157
5.16 Isobaric VLE of [emim][TfO] + H2O binary at 0.1013 MPa:
comparison of experimental data [24] shown in symbols with
eNRTL model results (Chemistry II): ( ), ( ) eNRTL
model................................................................................................................... 158
5.17 Activity coefficient of water of [emim][TfO] + H2O binary at
different temperatures: comparison of experimental data [23]
shown in symbols with eNRTL model results (Chemistry II)
shown as dotted lines: ( ), ( ) 323.3 K; ( ), ( )
333.3 K; ( ), ( ) 343.3 K ; ( ), ( ) 353.3 K; ( X
), ( ) 363.3 K. ........................................................................................... 159

xvi
Texas Tech University, Nazir Hossain, August 2018

CHAPTER 1
INTRODUCTION
1.1 Background
The information about phase equilibria of aqueous electrolytes is essential for many

industrial and natural processes such as desalination of seawater and “produced water”

from hydraulic fracturing [1-3], electro-dialysis, CO2 capturing and sequestration [4],

nuclear waste treatment [5-7], solubility phase diagram of salt-lake brine [8],

precipitation and crystallization processes in geothermal-energy systems, partitioning

processes in biological systems, etc. [9]. The addition of electrolytes to water or any other

solvent makes the system non-ideal in nature due to the presence of the ionic species. The

usual practice of representing the non-ideality is to optimize thermodynamic model

parameters by correlating available experimental data [9-14]. It is desirable that the

optimized model should predict the thermodynamic properties in a wide range of

concentration and temperature for the application of mass and energy balance in chemical

industries [15].

Thermodynamic modeling of aqueous electrolyte systems differs from that of

conventional non-electrolyte activity coefficient models since the long range ion-ion

interactions are crucial for modelling electrolyte systems [9]. There have been lot of

efforts to develop thermodynamic model for aqueous electrolyte systems. The most

commonly used activity coefficient models for aqueous electrolyte systems are the Pitzer

model [10], the extended UNIQUAC (EUNIQUAC) model [16, 17], the OLI-MSE model

[18, 19], ePC-SAFT [20], and electrolyte nonrandom two liquid (eNRTL) model [12, 13,

21]. Among them, the Pitzer ion-interaction model and eNRTL model the most widely
1
Texas Tech University, Nazir Hossain, August 2018

used thermodynamic activity coefficient model for the aqueous electrolyte systems. Both

models use the extended Pitzer-Debey-Huckel expression to account for the long range

electrostatic interactions [10, 12] . The Pitzer model uses a virial expansion or Margules

type ionic strength dependent equation for short range interactions. On the other hand,

eNRTL model adopted the Non Random Two Liquid theory to account for the short

range ion-ion, ion-water, and water-water interactions.

In Pitzer model, the excess Gibbs energy is expressed in a virial expansion type

equation, including many body interactions as a function of ionic strength. At a given

temperature, for a single aqueous electrolyte system, Pitzer model requires four

(0) (1) (2) 𝜙


adjustable parameters, 𝛽𝑀𝑋 , 𝛽𝑀𝑋 , 𝛽𝑀𝑋 , and 𝐶𝑀𝑋 , which are nonlinear functions of ionic

strength. The model parameters are optimized with the available experimental osmotic

and mean ionic activity coefficient data [9]. However, the Pitzer model is applicable only

up to ionic strength of10 molal [22]. The extrapolation of the Pitzer model beyond10

molal is not reliable due to the inherent limitation of virial expansion type equation [12]

and the nonlinear nature of the model parameters. Another reason of the failure of the

Pitzer model to extrapolate the mean ionic activity coefficient at high ionic strength is the

molality scale formulation of the excess Gibbs energy since the molality of the system

goes infinity as the mole fraction of the electrolytes approaches unity. Although the mole

fraction based Pitzer model [23] does not diverge at high concentration, the model

predicts very high activity coefficient at higher concentration which does not follow the

trend of the results from the supersaturation data and molecular dynamics simulation

[24].

2
Texas Tech University, Nazir Hossain, August 2018

The formulation of the eNRTL model is based on Non Random Two Liquid

theory where the excess Gibbs energy is the sum of the two contributions: long range and

short range. The long range interaction of the excess Gibbs energy is taken care with the

Pitzer-Debye-Hückel limiting law equation. The short range interactions are accounted

by the nonrandom two liquid theory. The two major assumptions for the short range

interactions are: a) like-ion repulsion and b) local electroneutrality of species around the

solvent molecule [12]. Requiring only two adjustable parameters, 𝜏𝑖𝑗 and 𝜏𝑗𝑖 , the eNRTL

model can correlate the experimental data accurately [12-14, 21]. Once the binary

interaction parameters are identified, the eNRTL model reliably extrapolates the

thermodynamic properties up to pure salt.

The fundamental property of interest for the aqueous electrolyte systems is the

mean ionic activity coefficient. All the other thermodynamic properties can essentially

derived from the mean ionic activity coefficient [9, 25]. So the particular interest is to

correlate the thermodynamic model parameters with the mean ionic activity coefficient

data, and to predict the mean ionic activity coefficient in a wide range of concentration

and temperature. For an example, for NaCl + H2O binary, both the Pitzer model (molality

and mole fraction based) and eNRTL model satisfactorily correlate the mean ionic

activity coefficient up to salt saturation concentration i.e, ionic strength around 6 molal at

298.15 K. However, the extrapolations of the mean ionic activity coefficient of Pitzer and

eNRTL diverge from each other beyond 6 molal. The molality based Pitzer model

predicts very high mean ionic activity coefficient and eventually diverges to infinity at

pure salt concentration. On the other hand, the eNRTL model predicts a gradual increase

3
Texas Tech University, Nazir Hossain, August 2018

in the mean ionic activity coefficient with increasing salt concentration and reaches

plateau. Although the mole fraction based Pitzer model does not diverge to infinity at

pure salt concentration it predicts a steep increase in mean ionic activity coefficient with

increasing salt concentration. Similar behaviors are observed for other aqueous

electrolyte systems such as magnesium chloride [26] and calcium chloride [27]. The

model predictions cannot be validated beyond the saturation concentration due to the

scarcity of the supersaturation data.

Molecular dynamics simulation is another independent tool to calculate the mean

ionic activity coefficient for the aqueous electrolyte systems [28-32]. There are two

methods for the calculation of the free energy of the aqueous electrolyte systems: direct

and indirect. In direct method, the free energy of the systems are calculated by the

particle insertion method [28]. The later method calculates free energy by using

Kirkwood- Buff theory where the structural property of system can be obtained from the

bulk properties [31, 33]. Both methods need reliable force field parameters for both ions

and water [29, 32]. The Kirkwood-Buff technique offers a simple methodology to

calculate the activity coefficient from the information of the liquid structure i.e. radial

distribution of the species in the system. The molecular dynamics simulation, either of

the method, offer an independent method to calculate the mean ionic activity coefficient

for virtually any concentration. So it is possible to simulate the system at very high

concentrations and the results can be compared with the supersaturation data (if

available) and the prediction from the Pitzer model and eNRTL model to justify the

applicability of the models at higher salt concentrations.

4
Texas Tech University, Nazir Hossain, August 2018

The thermodynamic models should be able to predict the thermodynamic

properties in the wide temperature range as well. The temperature dependence of the

interaction parameters of the Pitzer model is totally empirical and requires a lot of

temperature coefficients to fit the experimental data from 273.15-473.15 K temperature

[34]. For example, NaCl + H2O binary requires 36 adjustable parameters to correlate

experimental data from 273.15-473.15 K. The number of adjustable parameters increases

for the multicomponent aqueous electrolyte systems [22, 35]. From the process

simulation point , a large number of adjustable parameters make it cumbersome to

perform heat and mass balance [36]. On the other hand, the temperature dependence of

the interaction parameters of the eNRTL model is expressed as Gibbs-Helmholtz like

expression suggested by Clark and Glew [37]. This expression has three temperature

coefficients, 𝛥𝑔𝑖𝑗 , 𝛥ℎ𝑖𝑗 , and ∆𝑐𝑝 𝑖𝑗 which can be correlated with excess Gibbs energy,

enthalpy, and heat capacity, respectively, at 298.15 K [36] . The form of the expression of

the interaction parameters remain same for all electrolytes and can be correlated with the

experimental data. Example, for NaCl + H2O binary, six temperature coefficients are

enough to correlate and predict the experimental data up to pure salt and temperature

range from 273.15-473.15 K. For the multicomponent electrolyte systems, the minimal

amount of salt-water and salt-salt binary interaction parameters are enough to correlate

and predict thermodynamic properties in wide temperature and concentration range.

The ultimate goal is to develop an engineering thermodynamic model which can

correlate the experimental data and predict the thermodynamic properties of a

multicomponent aqueous electrolyte system in the wide concentration and temperature

5
Texas Tech University, Nazir Hossain, August 2018

range. Considering the minimal amount of adjustable parameters and proper

concentration and temperature dependence, the eNRTL model would be the ultimate

choice for modeling multicomponent aqueous electrolyte systems. There have been

extensive works are performe by Chen and co-workers to model the produced water [14,

26, 27, 38, 39]. Chen and co-workers are in the process of developing a comprehensive

thermodynamic modeling for the highly saline produced water to perform the mass and

energy balance in process simulation.

Since the eNRTL model can be extrapolate to the pure salt concentration it could

be an excellent candidate for the thermodynamic modeling of aqueous Ionic Liquids

(ILs). Ionic Liquids are molten salt having negligible vapor pressure which can be found

as liquid state at temperature less than 100 °C. Ionic Liquids are considered as “green

solvents,” and have gained a lot of attention for industrial applications, such as CO2

capturing [40-42], separation solvents [43], refrigerants, reaction media [44, 45], novel

electrolytes in batteries and photovoltaic cells [46] for their negligible vapor pressure

[47]. Most of the ILs are viscous, therefore, it is a common practice to use solvent such

as water to reduce the viscosity [48].

There have been extensive efforts on thermodynamic modeling of IL-containing

mixtures [49, 50], based on UNIQUAC [51], NRTL [52], eNRTL [12, 13], UNIFAC

[53], COSMO-RS [54], COSMO-SAC [55], and NRTL-SAC [56]. Simoni [49] reported

an extensive thermodynamic study(both experimental and modeling) on aqueous

solutions of 1-ethyl-3methylimidazolium ethylsulfate ([emim][EtSO4]), 1-ethyl-3-

methylimidazolium trifluoroacetate ([emim][TFA]), and 1-ethyl-3-methylimidazolium

6
Texas Tech University, Nazir Hossain, August 2018

trifluoromethanelsulfonate ([emim][TfO]) with the correlative NRTL and eNRTL

models. NRTL model considers, the ILs as molecular species, while in the eNRTL model

they are considered either completely dissociated (𝐶𝐴 → 𝐶 + + 𝐴− ) or fixed amount of

dissociation extent throughout the concentration. Lee and Lin [50] treated the IL

dissociations (𝐶𝐴 ↔ 𝐶 + + 𝐴− ) with chemical equilibrium assumption for the

[emim][EtSO4] + water binary and the [emim][TfO] + water binary with COSMO-SAC

model [55]. The [emim][TFA] + water binary was not included in their study. While the

model was able to achieve reasonable agreement with the dissociation extent data in

concentrated IL solutions, contrary to the nearly fixed dissciatiton extent observed for the

IL solutions at the infinite dilution of ILs in water [49], the model suggests a nearly

complete dissociation of ILs.

1.2 Content of the Thesis


In Chapter 2, the mean ionic activity coefficients of NaCl + H2O binary are

calculated from the molecular dynamics simulation in a wide range of concentration with

Kirkwood-Buff theory by considering finite size-effects [57]. We have chosen sodium

chloride as the representative electrolyte as it is the prominent component of seawater

and well-studied both experimentally [58-61] and with thermodynamic models [11, 14,

23]. Also, the Kirkwood-Buff force field parameters for this binary are well-established

[31] and are validated by free energy calculation method by Mester and Panagiotopoulos

[29]. The results from the molecular dynamics simulation are compared with the

supersaturation data and the prediction of the Pitzer model and eNRTL model. Our

7
Texas Tech University, Nazir Hossain, August 2018

results show that the mean ionic activity coefficient values calculated from molecular

simulations support the trend predicted by the eNRTL model.

In Chapter 3, the temperature dependence of the eNRTL model parameters are

expressed as a Gibbs-Helmholtz type equation. The three temperature coefficients of the

interaction parameters are correlated with excess Gibbs energy, excess enthalpy, and

excess heat capacity data at 298.15 K. The resulting parameters show the linear

temperature dependency and are able to reliably extrapolate the thermodynamic

properties at wide range of temperature.

In Chapter 4, the eNRTL model is applied for aqueous Li+–Na+–K+–Mg2+−SO42−

solutions quinary as an example of application of this model to a multicomponent

aqueous electrolyte system. It is an extension of our previous works aimed towards the

developing a comprehensive and accurate thermodynamic model for aqueous brine

solution. The eNRTL binary interaction parameters for (Li+ SO42−):H2O, (Li+ SO42−):(Na+

SO42−), (Li+ SO42−):(K+ SO42−), and (Li+ SO42−):(Mg2+ SO42−) pairs are obtained from the

correlation of the experimental osmotic coefficient, excess enthalpy, and solubility data.

The other binary interaction parameters are obtained from the literature [14, 26, 27, 38,

39, 62]. Additionally, solubility data are used to regress the thermodynamic constants for

the various solid salts. The model displays a significant predictive capability and

represents well the thermodynamic behavior of highly nonideal aqueous Li+–Na+–K+–

Mg2+–SO42– quinary system from infinite dilution up to saturation and temperatures from

273.15 K to 573.15 K.

8
Texas Tech University, Nazir Hossain, August 2018

In Chapter 5, the applicability of the symmetric eNRTL model is validated for

the aqueous Ionic Liquids system. In this chapter, the dissociation behavior of the three

ionic liquids (ILs): 1-ethyl-3methylimidazolium ethylsulfate ([emim][EtSO4]), 1-ethyl-3-

methylimidazolium trifluoroacetate ([emim][TFA]), and 1-ethyl-3-methylimidazolium

trifluoromethanelsulfonate ([emim][TfO]) in water are studied. To understand the issue

of the complete dissociation predicted by the eNRTL model and COSMO-SAC model at

infinite dilute concentrations of ILs, we perform molecular dynamics simulation of these

three ILs. Consistent with the literature experimental data showing these ILs are partially

dissociated, the molecular dynamics simulation shows the formation of singlets, doublets,

and triplets at very dilute concentration of ILs in water. Based on these findings, we

propose a dissociation chemistry for ILs and accurately correlate all available vapor

pressure and dissociation extent data over the entire IL concentration range for the three

ILs with symmetric electrolyte Non-Random Two Liquid model.

9
Texas Tech University, Nazir Hossain, August 2018

References
[1] K. V. Reddy and N. Ghaffour, "Overview of the Cost of Desalinated Water and
Costing Methodologies," Desalination, vol. 205, pp. 340-353, 2007.
[2] N. Ghaffour, T. M. Missimer, and G. L. Amy, "Technical Review and Evaluation
of The Economics of Water Desalination: Current and Future Challenges for
Better Water Supply Sustainability," Desalination, vol. 309, pp. 197-207, 2013.
[3] D. L. Shaffer, L. H. Arias Chavez, M. Ben-Sasson, S. Romero-Vargas Castrillón,
N. Y. Yip, and M. Elimelech, "Desalination and Reuse of High-Salinity Shale Gas
Produced Water: Drivers, Technologies, and Future Directions," Environmental
Science & Technology, vol. 47, pp. 9569-9583, 2013.
[4] S. M. Benson and F. M. Orr, "Carbon Dioxide Capture and Storage," MRS
Bulletin, vol. 33, pp. 303-305, 2011.
[5] M. Benedict, H. Levi, and T. Pigford, "Nuclear Chemical Engineering," 1982-12-
01 1982.
[6] D. D. Sood and S. K. Patil, "Chemistry of Nuclear Fuel Reprocessing: Current
status," Journal of Radioanalytical and Nuclear Chemistry, vol. 203, pp. 547-573,
1996.
[7] M. Wang, M. B. Gorensek, and C.-C. Chen, "Thermodynamic Representation of
Aqueous Sodium Nitrate and Nitric Acid Solution with Electrolyte NRTL
Model," Fluid Phase Equilibria, vol. 407, pp. 105-116, 2016.
[8] Y. Zeng, X. Lin, and X. Yu, "Study on the Solubility of the Aqueous Quaternary
System Li2SO4 + Na2SO4 + K2SO4 + H2O at 273.15 K," Journal of Chemical &
Engineering Data, vol. 57, pp. 3672-3676, 2012.
[9] J. M. Prausnitz, R. N. Lichtenthaler, and E. G. de Azevedo, Molecular
thermodynamics of fluid-phase equilibria: Pearson Education, 1998.
[10] K. S. Pitzer, "Thermodynamics of electrolytes. I. Theoretical basis and general
equations," The Journal of Physical Chemistry, vol. 77, pp. 268-277, 1973.
[11] K. S. Pitzer, J. C. Peiper, and R. H. Busey, "Thermodynamic Properties of
Aqueous Sodium Chloride Solutions," Journal of Physical and Chemical
Reference Data, vol. 13, pp. 1-102, 1984.
[12] C.-C. Chen, H. I. Britt, J. F. Boston, and L. B. Evans, "Local composition model
for excess Gibbs energy of electrolyte systems. Part I: Single solvent, single
completely dissociated electrolyte systems," AIChE Journal, vol. 28, pp. 588-596,
1982.
[13] C.-C. Chen and L. B. Evans, "A Local Composition Model for the Excess Gibbs
Energy of Aqueous Electrolyte Systems," AIChE Journal, vol. 32, pp. 444-454,
1986.
[14] Y. Yan and C.-C. Chen, "Thermodynamic Representation of the NaCl + Na2SO4
+ H2O System with Electrolyte NRTL Model," Fluid Phase Equilibria, vol. 306,
pp. 149-161, Jul 25 2011.
[15] S. H. Saravi, S. Honarparvar, and C.-C. Chen, "Modeling Aqueous Electrolyte
Systems," Chemical Engineering Progress, vol. 111, pp. 65-75, 2015.

10
Texas Tech University, Nazir Hossain, August 2018

[16] K. Thomsen, "Modeling Electrolyte Solutions with the Extended Universal


Quasichemical (UNIQUAC) Model," Pure and applied chemistry, vol. 77, pp.
531-542, 2005.
[17] F. F. Hingerl, T. Wagner, D. A. Kulik, K. Thomsen, and T. Driesner, "A New
Aqueous Activity Model for Geothermal Brines in the System Na-K-Ca-Mg-H-
Cl-SO4-H2O from 25 to 300 °C.," Chemical Geology, vol. 381, pp. 78-93, 2014.
[18] P. Wang, A. Anderko, and R. D. Young, "A Speciation-Based Model for Mixed-
Solvent Electrolyte Systems," Fluid Phase Equilibria, vol. 203, pp. 141-176,
2002.
[19] P. Wang, A. Anderko, R. D. Springer, and R. D. Young, "Modeling Phase
Equilibria and Speciation in Mixed-Solvent Electrolyte Systems: II. Liquid–
Liquid Equilibria and Properties of Associating Electrolyte Solutions," Journal of
Molecular Liquids, vol. 125, pp. 37-44, 2006.
[20] L. F. Cameretti, G. Sadowski, and J. M. Mollerup, "Modeling of Aqueous
Electrolyte Solutions with Perturbed-Chain Statistical Associated Fluid Theory,"
Industrial & Engineering Chemistry Research, vol. 44, pp. 3355-3362, 2005.
[21] Y. Song and C.-C. Chen, "Symmetric Electrolyte Nonrandom Two-Liquid
Activity Coefficient Model," Industrial & Engineering Chemistry Research, vol.
48, pp. 7788-7797, 2009.
[22] W. Voigt, "Chemistry of Salts in Aqueous Solutions: Applications, Experiments,
and Theory," in Pure and Applied Chemistry vol. 83, ed, 2011, p. 1015.
[23] S. L. Clegg and K. S. Pitzer, "Thermodynamics of multicomponent, miscible,
ionic solutions: generalized equations for symmetrical electrolytes," The Journal
of Physical Chemistry, vol. 96, pp. 3513-3520, 1992.
[24] N. Hossain, A. Ravichandran, R. Khare, and C. C. Chen, "Revisiting Electrolyte
Thermodynamic Models: Insights from Molecular Simulations," AIChE Journal,
2018.
[25] J. R. Elliott and C. T. Lira, Introductory chemical engineering thermodynamics
vol. 184: Prentice Hall PTR Upper Saddle River, NJ, 1999.
[26] S. Tanveer, H. Zhou, and C.-C. Chen, "Thermodynamic Model of Aqueous
Mg2+– Na+– K+–Cl− Quaternary system," Fluid Phase Equilibria, vol. 437, pp.
56-68, 2017.
[27] S. Tanveer and C.-C. Chen, "Thermodynamic Modeling of Aqueous Ca2+–Na+–
K+–Cl− Quaternary System," Fluid Phase Equilibria, vol. 409, pp. 193-206, 2016.
[28] Z. Mester and A. Z. Panagiotopoulos, "Mean Ionic Activity Coefficients in
Aqueous Nacl Solutions from Molecular Dynamics Simulations," The Journal of
Chemical Physics, vol. 142, p. 044507, 2015.
[29] Z. Mester and A. Z. Panagiotopoulos, "Temperature-Dependent Solubilities and
Mean Ionic Activity Coefficients of Alkali Halides in Water from Molecular
Dynamics Simulations," The Journal of Chemical Physics, vol. 143, p. 044505,
2015.
[30] H. Jiang, Z. Mester, O. A. Moultos, I. G. Economou, and A. Z. Panagiotopoulos,
"Thermodynamic and Transport Properties of H2O + NaCl from Polarizable Force

11
Texas Tech University, Nazir Hossain, August 2018

Fields," Journal of Chemical Theory and Computation, vol. 11, pp. 3802-3810,
2015.
[31] S. Weerasinghe and P. E. Smith, "A Kirkwood–Buff derived force field for
sodium chloride in water," The Journal of Chemical Physics, vol. 119, pp. 11342-
11349, 2003.
[32] M. B. Gee, N. R. Cox, Y. Jiao, N. Bentenitis, S. Weerasinghe, and P. E. Smith, "A
Kirkwood-Buff Derived Force Field for Aqueous Alkali Halides," Journal of
Chemical Theory and Computation, vol. 7, pp. 1369-1380, 2011.
[33] J. G. Kirkwood and F. P. Buff, "The Statistical Mechanical Theory of Solutions.
I," The Journal of chemical physics, vol. 19, pp. 774-777, 1951.
[34] K. S. Pitzer, Thermodynamics, 3rd ed. New York: McGraw Hill, 1995.
[35] W. Voigt, "Solubility Equilibria in Multicomponent Oceanic Salt Systems from t=
0 to 200 °C. Model Parameterization and Databases," Pure and applied chemistry,
vol. 73, pp. 831-844, 2001.
[36] N. Hossain, S. K. Bhattacharia, and C. C. Chen, "Temperature Dependence of
Interaction Parameters in Electrolyte NRTL Model," AIChE Journal, vol. 62, pp.
1244-1253, 2016.
[37] E. Clarke and D. Glew, "Evaluation of Thermodynamic Functions from
Equilibrium Constants," Transactions of the Faraday Society, vol. 62, pp. 539-
547, 1966.
[38] S. K. Bhattacharia and C.-C. Chen, "Thermodynamic Modeling of KCl + H2O and
KCl + NaCl + H2O Systems using Electrolyte NRTL Model," Fluid Phase
Equilibria, vol. 387, pp. 169-177, 2015.
[39] S. K. Bhattacharia, S. Tanveer, N. Hossain, and C.-C. Chen, "Thermodynamic
Modeling of Aqueous Na+–K+–Mg2+–SO42− Quaternary System," Fluid Phase
Equilibria, vol. 404, pp. 141-149, 2015.
[40] E. D. Bates, R. D. Mayton, I. Ntai, and J. H. Davis, "CO2 Capture by a Task-
Specific Ionic Liquid," Journal of the American Chemical Society, vol. 124, pp.
926-927, 2002.
[41] M. W. Arshad, "CO2 Capture Using Ionic Liquids," Technical University of
Denmark, DTU, DK-2800 Kgs. Lyngby, Denmark, 2009.
[42] M. Hasib-ur-Rahman, M. Siaj, and F. Larachi, "Ionic Liquids for CO2 Capture—
Development and Progress," Chemical Engineering and Processing: Process
Intensification, vol. 49, pp. 313-322, 2010.
[43] A. Chapeaux, L. D. Simoni, T. S. Ronan, M. A. Stadtherr, and J. F. Brennecke,
"Extraction of Alcohols from Water with 1-hexyl-3-methylimidazolium
bis(trifluoromethylsulfonyl)imide," Green Chemistry, vol. 10, pp. 1301-1306,
2008.
[44] K. Seddon, "Room‐Temperature Ionic Liquids: Neoteric Solvents for Clean
Catalysis," ChemInform, vol. 28, 1997.
[45] C. Reichardt, "Solvents and Solvent Effects:  An Introduction," Organic Process
Research & Development, vol. 11, pp. 105-113, 2007.
[46] K. Hayamizu, Y. Aihara, H. Nakagawa, T. Nukuda, and W. S. Price, "Ionic
Conduction and Ion Diffusion in Binary Room-Temperature Ionic Liquids

12
Texas Tech University, Nazir Hossain, August 2018

Composed of [emim][BF4] and LiBF4," The Journal of Physical Chemistry B, vol.


108, pp. 19527-19532, 2004.
[47] K. R. Seddon, "Ionic Liquids for Clean Technology," Journal of Chemical
Technology and Biotechnology, vol. 68, pp. 351-356, 1997.
[48] D. Constantinescu, K. Schaber, F. Agel, M. H. Klingele, and T. J. S. Schubert,
"Viscosities, Vapor Pressures, and Excess Enthalpies of Choline Lactate + Water,
Choline Glycolate + Water, and Choline Methanesulfonate + Water Systems,"
Journal of Chemical & Engineering Data, vol. 52, pp. 1280-1285, 2007.
[49] L. D. Simoni, Predictive modeling of fluid phase equilibria for systems containing
ionic liquids: University of Notre Dame, 2010.
[50] B.-S. Lee and S.-T. Lin, "A Priori Prediction of Dissociation Phenomena and
Phase Behaviors of Ionic Liquids," Industrial & Engineering Chemistry Research,
vol. 54, pp. 9005-9012, 2015.
[51] D. S. Abrams and J. M. Prausnitz, "Statistical Thermodynamics of Liquid
Mixtures: A New Expression for the Excess Gibbs Energy of Partly or
Completely Miscible Systems," AIChE Journal, vol. 21, pp. 116-128, 1975.
[52] H. Renon and J. M. Prausnitz, "Local Compositions in Thermodynamic Excess
Functions for Liquid Mixtures," AIChE journal, vol. 14, pp. 135-144, 1968.
[53] A. Fredenslund, R. L. Jones, and J. M. Prausnitz, "Group‐Contribution Estimation
of Activity Coefficients in Nonideal Liquid Mixtures," AIChE Journal, vol. 21,
pp. 1086-1099, 1975.
[54] A. Klamt, "Conductor-Like Screening Model for Real Solvents: a New Approach
to the Quantitative Calculation of Solvation Phenomena," The Journal of Physical
Chemistry, vol. 99, pp. 2224-2235, 1995.
[55] S.-T. Lin and S. I. Sandler, "A Priori Phase Equilibrium Prediction from a
Segment Contribution Solvation Model," Industrial & Engineering Chemistry
Research, vol. 41, pp. 899-913, 2002.
[56] C.-C. Chen and Y. Song, "Solubility Modeling with a Nonrandom Two-Liquid
Segment Activity Coefficient Model," Industrial & Engineering Chemistry
Research, vol. 43, pp. 8354-8362, 2004.
[57] R. Cortes-Huerto, K. Kremer, and R. Potestio, "Communication: Kirkwood-Buff
Integrals in The Thermodynamic Limit From Small-Sized Molecular Dynamics
Simulations," The Journal of Chemical Physics, vol. 145, p. 141103, 2016.
[58] R. A. Robinson and R. H. Stokes, Electrolyte Solutions: Courier Dover
Publications, 2002.
[59] W. J. Hamer and Y. C. Wu, "Osmotic Coefficients and Mean Activity
Coefficients of Uni‐univalent Electrolytes in Water at 25 °C," Journal of Physical
and Chemical Reference Data, vol. 1, pp. 1047-1100, 1972.
[60] I. N. Tang, H. R. Munkelwitz, and N. Wang, "Water Activity Measurements with
Single Suspended Droplets: The NaCl-H2O and KCl-H2O Systems," Journal of
Colloid and Interface Science, vol. 114, pp. 409-415, 1986.
[61] C. K. Chan, Z. Liang, J. Zheng, S. L. Clegg, and P. Brimblecombe,
"Thermodynamic Properties of Aqueous Aerosols to High Supersaturation: I—

13
Texas Tech University, Nazir Hossain, August 2018

Measurements of Water Activity of the System Na+−Cl−−NO3−−SO24−−H2O at ~


298.15 K," Aerosol Science and Technology, vol. 27, pp. 324-344, 1997.
[62] S. Tanveer and C.-C. Chen, "Molecular Thermodynamic Model for Aqueous
Na+–K+–Mg2+–Ca2+– SO42− System " In preperation, 2018.

14
Texas Tech University, Nazir Hossain, August 2018

CHAPTER 2
REVISITING ELECTROLYTE THERMODYNAMIC MODELS:
INSIGHTS FROM MOLECULAR SIMULATIONS
2.1 Abstract
Pitzer and electrolyte non-random two-liquid (eNRTL) models are the two most

widely used electrolyte thermodynamic models. For aqueous sodium chloride (NaCl)

solution, both models correlate the experimental mean ionic activity coefficient (𝛾± ) data

satisfactorily up to salt saturation concentration, i.e., ionic strength around 6 molal.

However, beyond 6 molal, the model extrapolations deviate significantly and diverge

from each other. We examine this divergence by calculating the mean ionic activity

coefficient over a wide range of concentration based on molecular simulations and

Kirkwood-Buff (KB) theory. We show that the asymptotic behavior of the activity

coefficient predicted by the eNRTL model is consistent with the molecular simulation

results and supersaturation experimental data.

Keywords: Pitzer model; eNRTL model; aqueous NaCl; Mean ionic activity coefficient;

Molecular dynamics simulation; Kirkwood-Buff theory

15
Texas Tech University, Nazir Hossain, August 2018

2.2 Introduction
Predicting the phase equilibrium and thermodynamic properties of aqueous

electrolyte solutions is important for designing industrial processes including hydraulic

fracturing [1, 2], desalination [3-5], nuclear waste treatment [6-8], among others.

Molecular thermodynamic models such as Pitzer model [9-11], electrolyte nonrandom

two liquid (eNRTL) model [12, 13], ePC-SAFT [14], etc., are often employed for

predicting the properties of these electrolyte solutions. One of the fundamental

thermodynamic properties that is of interest for modeling aqueous electrolyte solutions is

the mean ionic activity coefficient (𝛾±), which quantifies the non-ideality of the solution

phase. All of the other thermodynamic properties can be derived from the mean ionic

activity coefficient [15, 16]. Hence it is imperative for the electrolyte thermodynamic

models to predict mean ionic activity coefficients with a good accuracy.

The current practice in thermodynamic modeling of electrolyte solutions is to

correlate the experimental data for the activity coefficient and the other related

thermodynamic properties to optimize the model parameters, following which the model

is used to extrapolate for the prediction of 𝛾± over wider ranges of concentration and

temperature. Pitzer and eNRTL are the two most widely used electrolyte thermodynamic

models. For aqueous sodium chloride (NaCl) solution, Pitzer model (both molality [17]

and mole fraction-based [18]) and eNRTL model correlate the experimental mean ionic

activity coefficient data satisfactorily up to salt saturation concentration, i.e., ionic

strength around 6 molal. However, beyond 6 molal, the model extrapolations deviate

significantly and diverge from each other. While the 𝛾± predicted by the eNRTL model

shows a gradual increase with increasing salt concentration, the most widely used variant
16
Texas Tech University, Nazir Hossain, August 2018

of the Pitzer model (molality-based model) diverges to infinity beyond salt saturation

concentration. The mole fraction-based Pitzer model, although not diverging, shows a

much steeper change, and 𝛾± predicted by that model can be orders-of magnitude higher

than the prediction of eNRTL. From the above observations it is not clear which of the

models reliably predict the activity coefficient at higher concentration. This question

associated with the extrapolation of mean ionic activity coefficient to higher

concentration is also prevalent in other aqueous electrolyte solutions such as magnesium

chloride [19] and calcium chloride [20].

Molecular dynamics (MD) simulations have been extensively used to predict the

thermodynamic and phase equilibrium properties of systems involving fluids [21-24] and

solids [25]. With respect to the prediction of activity coefficients from molecular

simulations, two different classes of free energy techniques have been employed. The

first of these methods involves direct calculation of free energy [26] or chemical potential

[27, 28] from molecular simulations which are then used to calculate the mean ionic

activity coefficient. The approach of direct calculation of free energy using MD

simulations employs techniques such as thermodynamic integration [26] and particle

insertion [28]. The second method employs an indirect method where molecular

simulations are combined with Kirkwood-Buff (KB) theory [29] to calculate the chemical

potential and its derivatives such as activity coefficient [30-32]. The KB theory utilizes

the knowledge of the liquid structure to arrive at the free energy of the system. Extensive

work by Panagiotopoulos and co-workers [27, 28, 33] and Smith and co-workers [30-32],

using these methods, showed that the calculations of mean ionic activity coefficient from

17
Texas Tech University, Nazir Hossain, August 2018

molecular simulations are sensitive to the force field parameters used. Both the

techniques (direct free energy calculations and KB theory) can predict mean ionic activity

coefficient with good accuracy provided a good quality force field [27, 31] is available.

We note that the above-mentioned techniques have only been used to calculate 𝛾± of

NaCl up to salt saturation [27, 30, 31], i.e., ionic strength around 6 molal.

In this work, we use MD simulations to calculate the mean ionic activity

coefficient of aqueous sodium chloride over a wide range of concentration (below and

above the saturation concentration of NaCl) with the Kirkwood-Buff theory. These

results are used to examine the divergence between the predictions of the Pitzer and the

eNRTL models. We have chosen sodium chloride as the representative electrolyte as it is

the prominent component of seawater and well-studied both experimentally [34-37] and

with thermodynamic models [5, 17, 18]. MD simulations offer an independent technique

that can validate the predictions of these thermodynamic models at high concentration.

Our results show that the mean ionic activity coefficient values calculated from molecular

simulations support the trend predicted by the eNRTL model.

Rest of the article is organized as follows. Brief theoretical discussion and the

computation methodology used for the calculation of activity coefficient is provided in

Theoretical and Computational Details section, while the calculated activity coefficients

along with the discussion of results are presented in Results and Discussion section. The

conclusions are followed in Discussion and Conclusions section.

18
Texas Tech University, Nazir Hossain, August 2018

2.3 Theoretical and Computational Details

2.3.1 Theoretical Background


In this work, we use Kirkwood-Buff theory [29] (KB theory) to calculate the mean ionic

activity coefficient of aqueous NaCl. The choice of method between the KB theory and

direct free energy techniques to calculate the activity coefficient is dictated by the ease of

use. We have chosen to use KB theory since the computational implementation of the

technique is relatively straightforward.

KB theory relates the molecular structure of the liquid to the thermodynamic

properties such as chemical potential and mean ionic activity coefficient [29, 31] through

the Kirkwood-Buff Integral (KBI). Mathematically for an isotropic system, KBI is given

by:
∞ 𝜇𝑉𝑇
𝐺𝑖𝑗∞ = 4𝜋 ∫0 [ 𝑔𝑖𝑗 (𝑟) − 1] 𝑟 2 𝑑𝑟 (2.1)

Here, 𝐺𝑖𝑗∞ is the KB integral between components 𝑖 and 𝑗. The superscript on 𝐺𝑖𝑗 denotes

𝜇𝑉𝑇
that the quantity is evaluated in an open system. 𝑔𝑖𝑗 (𝑟) is the radial distribution

function (RDF) between components 𝑖 and 𝑗 in the grand canonical ensemble (constant

chemical potential, volume, and temperature ensemble). Following the definition of KBI,

the activity coefficient of the components can be calculated using [31, 38]:

𝜕𝑙𝑛 𝛾 1
𝑎𝑖𝑖 = 1 + (𝜕𝑙𝑛 𝜌𝑖 ) = ∞ −𝐺 ∞ )
(2.2)
𝑖 𝑃,𝑇 1+𝜌𝑖 (𝐺𝑖𝑖 𝑖𝑗

where 𝛾𝑖 is the activity coefficient of the ionic or molecular species on the molarity scale

and 𝜌𝑖 is the corresponding total number density of the ionic or molecular species in the

solution. 𝐺𝑖𝑖∞ and 𝐺𝑖𝑗∞ are the KB integrals between the components 𝑖𝑖 and 𝑖𝑗,

respectively, as defined in Eq. 2.1.

19
Texas Tech University, Nazir Hossain, August 2018

Note that according to Eq. 2.1, 𝐺𝑖𝑖∞ and 𝐺𝑖𝑗∞ should be evaluated in an open

system. However, the common practice in simulations is to replace the upper limit of the

integral in Eq. 2.1 by a finite value, 𝑅, and calculate the KB integrals from the

corresponding closed system [30, 32]. For such calculations, it is crucial to ensure

convergence of the integral in Eq. 2.1. This is usually accomplished by choosing a value

of 𝑅 that is much larger than the largest correlation length scale in the system [39, 40].

Such an approach becomes intractable in the presence of long range correlations in the

simulation, especially in the presence of ionic species, thereby demanding simulation of

systems with a very large number of atoms.

To overcome these problems associated with the convergence of KB integral in a

closed system, we follow the technique proposed by Cortes-Huerto et al. [39]. The

method involves obtaining the value of the KB integral of an open system (𝐺𝑖𝑗∞ ) from that

of a closed system (𝐺𝑖𝑗 ). Consider a simulation system of volume 𝑉0 in an isothermal-

isobaric ensemble (constant number of particles, pressure, and temperature ensemble)

which can be randomly divided into 𝑘 sub-domains each of volume 𝑉. Note that the 𝑘

sub-domains are not unique and there can be spatial overlap between these hypothetical

domains. Each of these sub-domains can now be considered as a finite simulation system

which can exchange particles with the bath, thereby mimicking an open system (𝜇𝑉𝑇

ensemble). The KB integral in each of these sub-domains can be calculated as [29, 39]:

〈𝑁𝑖 𝑁𝑗 〉−〈𝑁𝑖 〉〈𝑁𝑗 〉 𝛿


𝐺𝑖𝑗 (𝑉, 𝑉0 ) = 𝑉 ( 〈𝑁𝑖 〉〈𝑁𝑗 〉
− 〈𝑁𝑖𝑗〉) (2.3)
𝑖

20
Texas Tech University, Nazir Hossain, August 2018

Here 𝑁𝑖 and 𝑁𝑗 are the number of particles in the sub-domain of volume 𝑉. 𝛿𝑖𝑗 represents

the Kronecker delta. Cortes-Huerto et al. related 𝐺𝑖𝑗∞ to 𝐺𝑖𝑗 by including finite-size effects

[39]:

𝛿𝑖𝑗 𝛼𝑖𝑗
𝜆𝐺𝑖𝑗 (𝑉, 𝑉0 ) = 𝜆𝐺𝑖𝑗∞ (1 − 𝜆3 ) − 𝜆3 + 1/3 (2.4)
𝜌𝑖 𝑉0

1
𝑉 3
where 𝜆 ≡ (𝑉 ) , 𝜌𝑖 is the number density of species 𝑖, and 𝛼𝑖𝑗 is a constant that describes
0

the correlation between the atoms inside and outside the boundary of the sub-domain 𝑉.

For small values of 𝜆, Eq. 2.4 is linear in 𝜆𝐺𝑖𝑗 and the value of 𝐺𝑖𝑗∞ can be obtained from

the slope of the linear regime in the plot of 𝜆𝐺𝑖𝑗 vs 𝜆.

MD simulations were performed to obtain the values of the KB integral according

to Eq. 2.4. These KBI values were further used along with Eq. 2.2 to calculate the

derivatives of the activity coefficients (in the form of 𝑎𝑖𝑖 ). Finally, the activity coefficient

(𝛾𝑖 ) was calculated by numerically integrating 𝑎𝑖𝑖 . The activity coefficients thus obtained

on the molarity scale were converted to the required values on the concentration scale

using appropriate conversion expressions [15]. The following notations are consistently

used throughout this article. The mean ionic activity coefficient of the salt is denoted as

𝛾± and the activity coefficient of water is represented as 𝛾𝑤 . The superscripts on the

activity coefficients denote either the molarity (𝑀), molality (𝑚), or the mole fraction (𝑥)

scale activity coefficients.

2.3.2 Simulation Details


Molecular dynamics simulations of water-sodium chloride mixtures of different

concentrations of the salt (0.5 molal to 13 molal) were performed in the constant NPT

21
Texas Tech University, Nazir Hossain, August 2018

ensemble. In all of the simulations, water molecules were represented using the SPC/E

model [41] and the ions were described with the Kirkwood-Buff force field [30, 31]

(KBFF). For completeness, the force field parameters used in the simulations are listed in

Table 2.1. All of the simulations were performed using the LAMMPS simulation

package [42] at temperature 298.15 K and pressure 1 bar. Nosé-Hoover thermostat and

barostat [43] were used to maintain constant temperature and pressure. All interactions

were truncated at 9 Å. While the long range part of the electrostatic interactions were

handled by the particle-particle particle-mesh algorithm (PPPM) [44], tail corrections

were applied to van der Waals interactions. Each simulation system consisted of 22,000

water molecules with the number of ions varying as per the concentration of the mixture.

For clarity, the number of ions used for each concentration is provided in Table 2.2. The

simulations were initialized by randomly placing water molecules and ions at a density

close to that of liquid water following which the overlap in the initial configurations was

removed by applying soft potentials. All simulations in the NPT ensemble were carried

out for 50 ns duration, of which the first 10 ns of the trajectory were discarded for initial

equilibration. The remaining part of the trajectories were used for the calculation of the

KB integral. We note that, even at high salt concentration, the orientational relaxation

time for water molecules is much less than 25 ps (determined from the orientational time

correlation function of the O-H bond vector of water). Hence 10 ns of initial runs are

sufficient to attain equilibrium. Furthermore, we found that the water-water RDF

averaged over different time intervals (10-14 ns and 46-50 ns) of the simulations were

identical, confirming that the simulation had attained equilibrium. The simulation

22
Texas Tech University, Nazir Hossain, August 2018

trajectories were divided into ten blocks of 4 ns length and the KB integral was evaluated

in each block by considering 1000 randomly placed sub-domains (𝑘) at each value of the

volume 𝑉. Further, to improve the statistics in the calculation of the activity coefficient,

three replicas of each system were simulated. Finally, the mean values of the activity

coefficient and the standard deviation in the quantities were determined using the

bootstrapping technique [45].

2.3.3 Thermodynamic Modeling


Molality-based Pitzer and eNRTL models were used to predict the mean ionic

activity coefficients and water activity coefficients of aqueous NaCl at 298.15 K and 1

bar using ASPEN V8.4 software [46]. The model parameters were taken from the papers

of Pitzer et al. [17] and Yan and Chen [5], for the Pitzer and the eNRTL models,

respectively. These model parameters were obtained by fitting the experimental data up

to salt saturation concentration (~6 molal). The mole fraction-based Pitzer model

expressions were adopted from Clegg and Pitzer [18] which comprise an extended

Debye-Hückel term plus a Margules expansion carried out to the four suffix level.

Furthermore, the model parameters have been optimized with the experimental data

slightly above saturation (~7 molal) [18].

2.4 Results and Discussion

2.4.1 Mean Ionic Activity Coefficient Predicted by the eNRTL Model


Qualitatively Agrees with MD Calculations and Experimental Data
The computational procedure adopted for obtaining 𝐺𝑖𝑗∞ , specified in Eq. 2.4, is

illustrated in Figure 2.1 using the 𝜆𝐺𝑖𝑗 vs 𝜆 plot at 12 molal NaCl concentration. As

shown in the figure, the linear region in the plot was identified up to λ=0.3 and the values
23
Texas Tech University, Nazir Hossain, August 2018

of the KB integrals (𝐺𝑖𝑗∞ ) obtained were used along with Eq. 2.2 to predict the activity

coefficients. We note that a similar procedure was followed for all concentrations to

calculate the KB integrals, expect at low concentrations (0.56 - 3 molal) where the linear

region was identified up to λ=0.1.

Figures 2.2 and 2.3 show the MD simulation results for the logarithm of mean

ionic activity coefficient of NaCl (𝛾±) on the mole fraction and molality scale,

respectively, at temperature of 298.15 K and at 1 bar pressure along with experimental

data [36] and the predictions of the Pitzer and the eNRTL models. The experimental

mean ionic activity coefficient data are smoothed data calculated from the measured

water activity data through Gibbs-Duhem equation [36]. The values of activity coefficient

on different scales (molarity, molality, and mole fraction) calculated from MD

simulations are also reported in Table 2.3. Since the behavior of the activity coefficient at

high salt concentration (beyond saturation) is of primary interest in the present work,

simulations at very low concentrations were not performed. We also note that, calculating

the mean ionic activity coefficient at very low salt concentration using the present

methodology requires systems of large size to avoid statistical fluctuations. Such

calculations are computationally demanding and are not considered in this work.

The ability of the KB theory to predict 𝛾± was assessed by comparing its

predictions with experimental data [34] and the calculations of Weerasinghe and Smith

[30] in the concentration range of 0.5 – 6 molal. It can be seen from Figure 2.2 that,

within the limits of statistical uncertainties, predictions of KB theory (this work) show a

good agreement with the molecular simulations results of Weerasinghe and Smith [30,

24
Texas Tech University, Nazir Hossain, August 2018

31] (that were obtained by integrating the radial distribution function). However, the

results from both (ours and Weerasinghe and Smith) molecular dynamics simulation

studies do not show exact agreement with the experimental data and the model

predictions within the saturation limit. We note that the calculations of mean ionic

activity coefficients from simulations are sensitive to the details of the water and ion

force field parameters used [27, 28, 33]. The combination of SPC/E and KBFF was used

as the force field in the present work as it provides the best prediction of 𝛾± compared

with experimental data as shown previously by Mester and Panagiotopoulos [27].

Systematic study of the effect of water and ion force fields for predicting the mean ionic

activity coefficient is beyond the scope of this work.

The mean ionic activity coefficients calculated from simulations qualitatively

follow the trend predicted by the eNRTL model (Figure 2.2 and 2.3). Note that, unlike

𝛾± expressed on the mole fraction scale, the molality scale mean ionic activity coefficient

value (both from simulation and eNRTL model) shown in Figure 2.3 exhibits a

maximum at high concentration. Such a behavior is an artifact of the molality scale and

has no physical significance. Beyond the saturation concentration of sodium chloride (~6

molal), both eNRTL and MD simulations predict a gradual increase in the values of the

(𝑥)
logarithm of 𝛾± with increasing concentration, eventually exhibiting a plateau.

Consistent with this observation, experimental data [36] also shows a gradual increase in

(𝑥)
logarithm of 𝛾± with increasing concentration of NaCl and exhibit signs of reaching a

plateau. On the other hand, the predictions of the Pitzer models (both mole fraction and

(𝑥)
molality-based) show steeper increase in values of mean ionic activity coefficients, 𝛾± .

25
Texas Tech University, Nazir Hossain, August 2018

With the model parameters fitted to experimental data [36] above salt saturation (~7

molal), the mole fraction-based Pitzer model exhibits asymptotic behavior at much higher

salt concentration and the magnitude of the mean ionic activity coefficients predicted by

this model are also significantly higher when compared with experiments. The prediction

from molality-based Pitzer model eventually diverges at higher salt concentration as the

molality scale tends to infinity as NaCl concentration approaches unity. Because of this

artifact associated with the molality-based Pitzer model, in what follows, we focus on the

mole fraction-based Pitzer model to compare our results.

(𝑥)
For clarity, Figure 2.4 shows the prediction of the logarithm of 𝛾± by the

eNRTL model, beyond the concentrations presented in Figure 2.2. Both the local and

the Pitzer-Debye-Hückel contributions to the logarithm of mean ionic activity coefficient

attain a constant value as the salt concentration approach unity. The values of the mean

ionic activity coefficient predicted by the mole fraction-based Pitzer model is also shown

(𝑥)
in Figure 2.4. As discussed previously, 𝛾± predicted by the mole fraction-based Pitzer

model is significantly higher when compared to the eNRTL model predictions. To further

explain the difference between the models, Figure 2.5 shows the logarithm of the

(𝑥)
apparent mole fraction-based water activity coefficient, 𝛾𝑤 , predicted using the models

(𝑥)
and MD simulations, along with experimental data [36, 37]. The 𝛾𝑤 predicted by the

Pitzer model (both molality and mole fraction-based) shows a steep change at high NaCl

concentration and diverges from experimental data. On the other hand, as in the case for

(𝑥)
the mean ionic activity coefficient of the salt, the 𝛾𝑤 of water predicted by the eNRTL

model is in qualitative agreement with MD simulations and experimental data. They are
26
Texas Tech University, Nazir Hossain, August 2018

show linear trends at high NaCl concentration. A similar conclusion can also be drawn

from water activity data compared in Figure 2.6. Both molality and mole fraction-based

Pitzer model predictions of water activity show sharp decrease in comparison with

experimental data [36, 37] while the eNRTL predictions and MD simulations show the

qualitative agreement with experimental trend at high NaCl concentration. Hence,

following Gibbs-Duhem equation, we attribute the inability of Pitzer model to reliably

extrapolate 𝛾± beyond ionic strength of 6 molal to its inability to predict the water

activity coefficient [19].

Though the values of the activity coefficients calculated using MD simulations

and the eNRTL model are in qualitative agreement, there is a significant quantitative

disagreement between the two approaches. The values of mean ionic activity coefficient

calculated using MD simulations are higher than the experimental data at concentration

less than 6 molal and lower than the predictions of the eNRTL model and experimental

data at higher concentration. While the KB force field parameters should be further

optimized, additional comments on the MD simulations are noted here. At high

concentration of the salt, electrostatic screening between the ion pairs becomes important

and the eNRTL model implicitly takes that into account through the Pitzer-Debye-Hückel

formulation [12, 13]. However, such screening effects are not explicitly accounted for in

MD simulations and as a result the cation-anion binding (ion paring) energy is generally

overestimated [47] by classical MD force fields. Specifically, the force field used in this

work (KBFF) is parameterized at moderate salt concentrations and its application at high

salt concentrations should be accompanied by explicit accounting of ion pair screening.

27
Texas Tech University, Nazir Hossain, August 2018

Stronger net interaction between cation-anion pair results in the underestimation of

osmotic coefficient [47] and consequently the mean ionic activity coefficient is

underestimated when compared to the eNRTL model. A similar reasoning as explained

above can also be applied to understand the quantitative difference between MD

simulations and supersaturation experimental data. Nevertheless, following the

qualitative agreement in the trend of the activity coefficients between MD calculations,

experimental data, and the eNRTL model, we conclude that the asymptotic behavior of

mean ionic activity coefficient is consistently predicted by the eNRTL model as opposed

to the Pitzer model.

2.5 Discussion and Conclusions


We examined the extrapolations above 6 molal concentration of the mean ionic

activity coefficient of aqueous sodium chloride as predicted by Pitzer and eNRTL

models, the two widely used thermodynamic formalisms for treating aqueous electrolyte

solutions. Molecular simulation results and experimental data show that the asymptotic

behavior of 𝛾± at high salt concentration is consistent with the prediction of the eNRTL

model and in disagreement with the Pitzer model which predicts a rapid change of 𝛾± at

high salt concentration. Specifically, the widely used molality-based Pitzer model

diverges at high concentration due its formulation based on the molality scale that tends

to infinity as salt concentration approaches unity.

The difference between the e-NRTL model and the Pitzer model can be explained

as follows. In the limit of very high salt concentration local van der Waals interactions

and the electrostatic screening of the ions pairs become significant. Both of these models

28
Texas Tech University, Nazir Hossain, August 2018

describe the electrostatic contribution to the free energy (and consequently the mean ionic

activity coefficient) using the modified form of the Debye-Hückel theory. However, the

local interactions in the Pitzer model are represented using a virial-like or Margules

expansion which varies rapidly at very high concentration of the salt. The eNRTL model

on the other hand, describes the molecular interactions in terms of the local composition

approach which exhibits the correct asymptotic behavior. Furthermore, both the molality-

based and mole fraction-based Pitzer models incorporate empirical functions for both

ionic strength dependence and temperature dependence which require a large number of

adjustable parameters as input (36 parameters for the molality-based model [17] for

temperature from 273 to 573 K; 4 parameters for the mole fraction-based model [18] at

298.15 K). In contrast, the eNRTL model requires a smaller set of adjustable parameters

[5] (6 parameters for temperature from 273 to 473 K) and it offers more reliable

extrapolations at high salt concentration.

29
Texas Tech University, Nazir Hossain, August 2018

Tables

Table 2.1 Kirkwood-Buff derived force field parameters for NaCl [30] and SPC/E
parameters for water [41].

Model Atom/Ion 𝜖 (kJ/mol) 𝜎 (nm) 𝑞 (|𝑒|)

Na+ 0.320 0.245 +1.000


KBFF
Cl- 0.470 0.440 -1.000

O 0.650 0.317 -0.848


SPC/E
H 0.000 0.000 +0.424

For ions, geometric mixing rules were applied for both 𝜖𝑖𝑗 and 𝜎𝑖𝑗 as suggested by the
Kirkwood-Buff force field[30]. For Na+ and O, a scaling factor 0.75 was used for 𝜖𝑖𝑗 .

30
Texas Tech University, Nazir Hossain, August 2018

Table 2.2 Number of NaCl and water molecules at different concentrations used in
molecular dynamics simulations.

Molality Mole fraction 𝑁𝑤𝑎𝑡𝑒𝑟 𝑁𝑁𝑎𝐶𝑙

0.56 0.010 22000 222

1 0.018 22000 396

2 0.035 22000 793

3 0.051 22000 1189

4 0.067 22000 1585

5 0.083 22000 1982

6 0.098 22000 2378

7 0.112 22000 2774

8 0.126 22000 3171

9 0.140 22000 3567

10 0.153 22000 3963

11 0.165 22000 4360

12 0.178 22000 4756

13 0.190 22000 5152

31
Texas Tech University, Nazir Hossain, August 2018

Table 2.3 The values of activity coefficient of salt and water on different scales predicted
by MD simulations. The superscripts 𝑀, 𝑚, and 𝑥 denote molarity, molality, and mole
fraction scales, respectively. The water activity coefficient is based on the apparent mole
fraction of water.

(𝑀) (𝑚) (𝑥) (𝑥)


Molality 𝑙𝑛𝛾± 𝑙𝑛𝛾± 𝑙𝑛𝛾± 𝑙𝑛𝛾𝑤

0.56 -0.3926±0.0005 -0.3991±0.0006 -0.3792±0.0005 0.0075±0.0005

1 -0.3661±0.0031 -0.3797±0.0032 -0.3442±0.0031 -0.0132±0.0005

2 -0.2836±0.0063 -0.3143±0.0063 -0.2444±0.0064 -0.0278±0.0006

3 -0.1970±0.0075 -0.2476±0.0076 -0.1448±0.0075 -0.0475±0.0007

4 -0.1067±0.0155 -0.1776±0.0156 -0.0436±0.0145 -0.0730±0.0015

5 -0.0224±0.0282 -0.1144±0.0284 0.0484±0.0264 -0.1022±0.0036

6 0.0448±0.0364 -0.0698±0.0371 0.1209±0.0350 -0.1310±0.0057

7 0.0916±0.0390 -0.0468±0.0410 0.1727±0.0393 -0.1565±0.0066

8 0.1240±0.0397 -0.0391±0.0431 0.2092±0.0415 -0.1791±0.0074

9 0.1477±0.0413 -0.0405±0.0438 0.2354±0.0427 -0.1994±0.0080

10 0.1640±0.0413 -0.0491±0.0442 0.2536±0.0433 -0.2174±0.0085

11 0.1746±0.0412 -0.0620±0.0451 0.2664±0.0442 -0.2333±0.0088

12 0.1797±0.0417 -0.0800±0.0459 0.2734±0.0454 -0.2468±0.0095

32
Texas Tech University, Nazir Hossain, August 2018

Table 2.3 Continued

13 0.1818±0.0424 -0.1008±0.0464 0.2777±0.0457 -0.2588±0.0099

33
Texas Tech University, Nazir Hossain, August 2018

Figures

Figure 2.1 Procedure for obtaining the value of Kirkwood-Buff integral is illustrated at
12 molal NaCl concentration. The dashed black line indicates the region up to which the
linear fit was performed to calculateGij∞ .

34
Texas Tech University, Nazir Hossain, August 2018

Figure 2.2 Mean ionic activity coefficient of NaCl as a function of salt concentration at
298.15 K and 1 bar on mole fraction scale. The predictions from Pitzer [17, 18] and
eNRTL [5] models are compared with that calculated from KB theory and experimental
data (Robinson and Stokes [34] and Tang et al. [36]). 𝛾± calculations of Weerasinghe and
Smith [30] are also shown.

35
Texas Tech University, Nazir Hossain, August 2018

Figure 2.3 Mean ionic activity coefficient of NaCl as a function of salt concentration at
298.15 K and 1 bar on molality scale. The predictions from Pitzer [17, 18] and eNRTL
[5] models are compared with that calculated from KB theory and experimental data
(Robinson and Stokes [34] and Tang et al. [36]). 𝛾± calculations of Weerasinghe and
Smith [30] are also shown.

36
Texas Tech University, Nazir Hossain, August 2018

Figure 2.4 Local and the electrostatic contribution to the mean ionic activity coefficient
of NaCl over a wide range of salt concentrations as predicted by the eNRTL model. For
comparison, the activity coefficient predicted by the Pitzer model is also shown.

37
Texas Tech University, Nazir Hossain, August 2018

Figure 2.5 Comparison of water activity coefficient as a function of salt concentration at


298.15 K and 1 bar as predicted by the Pitzer and the eNRTL models. Activity coefficient
values of water predicted using KB theory are also shown in red symbols and the
experimental measurements of Tang et al. [36] and Chan et al. [37] are shown in black
symbols.

38
Texas Tech University, Nazir Hossain, August 2018

Figure 2.6 Comparison of water activity as a function of salt concentration at 298.15 K


and 1 bar as predicted by the Pitzer and the eNRTL models. The activity values of water
predicted using KB theory are also shown in red symbols and the experimental
measurements of Tang et al.[36] and Chan et al. [37] are shown in black symbols.

39
Texas Tech University, Nazir Hossain, August 2018

References
[1] D. L. Shaffer, L. H. Arias Chavez, M. Ben-Sasson, S. Romero-Vargas Castrillón,
N. Y. Yip, and M. Elimelech, "Desalination and Reuse of High-Salinity Shale Gas
Produced Water: Drivers, Technologies, and Future Directions," Environmental
Science & Technology, vol. 47, pp. 9569-9583, 2013.
[2] S. Honarparvar, S. H. Saravi, D. Reible, and C.-C. Chen, "Comprehensive
Thermodynamic Modeling Of Saline Water With Electrolyte NRTL Model: A
Study on Aqueous Ba2+-Na+-Cl−-SO42− Quaternary system," Fluid Phase
Equilibria, vol. 447, pp. 29-38, 2017.
[3] K. V. Reddy and N. Ghaffour, "Overview of the Cost of Desalinated Water and
Costing Methodologies," Desalination, vol. 205, pp. 340-353, 2007.
[4] N. Ghaffour, T. M. Missimer, and G. L. Amy, "Technical Review and Evaluation
of The Economics of Water Desalination: Current and Future Challenges for
Better Water Supply Sustainability," Desalination, vol. 309, pp. 197-207, 2013.
[5] Y. Yan and C.-C. Chen, "Thermodynamic Representation of the NaCl + Na2SO4
+ H2O System with Electrolyte NRTL Model," Fluid Phase Equilibria, vol. 306,
pp. 149-161, Jul 25 2011.
[6] M. Benedict, H. Levi, and T. Pigford, "Nuclear Chemical Engineering," 1982-12-
01 1982.
[7] D. D. Sood and S. K. Patil, "Chemistry of Nuclear Fuel Reprocessing: Current
Status," Journal of Radioanalytical and Nuclear Chemistry, vol. 203, pp. 547-
573, 1996.
[8] M. Wang, M. B. Gorensek, and C.-C. Chen, "Thermodynamic Representation of
Aqueous Sodium Nitrate and Nitric Acid Solution with Electrolyte NRTL
Model," Fluid Phase Equilibria, vol. 407, pp. 105-116, 2016.
[9] K. S. Pitzer, "Thermodynamics of electrolytes. I. Theoretical Basis and General
Equations," The Journal of Physical Chemistry, vol. 77, pp. 268-277, 1973.
[10] K. S. Pitzer and G. Mayorga, "Thermodynamics of Electrolytes. II. Activity and
Osmotic Coefficients for Strong Electrolytes with One or Both Ions Univalent,"
The Journal of Physical Chemistry, vol. 77, pp. 2300-2308, 1973.
[11] K. S. Pitzer, "The Treatment of Ionic Solutions over the Entire Miscibility
Range," Berichte der Bunsengesellschaft für physikalische Chemie, vol. 85, pp.
952-959, 1981.
[12] C.-C. Chen, H. I. Britt, J. F. Boston, and L. B. Evans, "Local Composition Model
for Excess Gibbs Energy of Electrolyte Systems. Part I: Single Solvent, Single
Completely Dissociated Electrolyte Systems," AIChE Journal, vol. 28, pp. 588-
596, 1982.
[13] Y. Song and C.-C. Chen, "Symmetric Electrolyte Nonrandom Two-Liquid
Activity Coefficient Model," Industrial & Engineering Chemistry Research, vol.
48, pp. 7788-7797, 2009.
[14] L. F. Cameretti, G. Sadowski, and J. M. Mollerup, "Modeling of Aqueous
Electrolyte Solutions with Perturbed-Chain Statistical Associated Fluid Theory,"
Industrial & Engineering Chemistry Research, vol. 44, pp. 3355-3362, 2005.

40
Texas Tech University, Nazir Hossain, August 2018

[15] J. M. Prausnitz, R. N. Lichtenthaler, and E. G. de Azevedo, Molecular


thermodynamics of fluid-phase equilibria: Pearson Education, 1998.
[16] J. R. Elliott and C. T. Lira, Introductory Chemical Engineering Thermodynamics
vol. 184: Prentice Hall PTR Upper Saddle River, NJ, 1999.
[17] K. S. Pitzer, J. C. Peiper, and R. H. Busey, "Thermodynamic Properties of
Aqueous Sodium Chloride Solutions," Journal of Physical and Chemical
Reference Data, vol. 13, pp. 1-102, 1984.
[18] S. L. Clegg and K. S. Pitzer, "Thermodynamics of Multicomponent, Miscible,
Ionic Solutions: Generalized Equations for Symmetrical Electrolytes," The
Journal of Physical Chemistry, vol. 96, pp. 3513-3520, 1992.
[19] S. Tanveer, H. Zhou, and C.-C. Chen, "Thermodynamic Model of Aqueous
Mg2+–Na+– K+–Cl− Quaternary System," Fluid Phase Equilibria, vol. 437, pp. 56-
68, 2017.
[20] S. Tanveer and C.-C. Chen, "Thermodynamic modeling of aqueous Ca2+–Na+–
K+–Cl− quaternary system," Fluid Phase Equilibria, vol. 409, pp. 193-206, 2016.
[21] J. J. Potoff and J. I. Siepmann, "Vapor–Liquid Equilibria of Mixtures Containing
Alkanes, Carbon Dioxide, And Nitrogen," AIChE Journal, vol. 47, pp. 1676-
1682, 2001.
[22] R. Khare, A. K. Sum, S. K. Nath, and J. J. de Pablo, "Simulation of Vapor−Liquid
Phase Equilibria of Primary Alcohols and Alcohol−Alkane Mixtures," The
Journal of Physical Chemistry B, vol. 108, pp. 10071-10076, 2004.
[23] A. Z. Panagiotopoulos, "Direct Determination of Phase Coexistence Properties of
Fluids By Monte Carlo Simulation in a New Ensemble," Molecular Physics, vol.
61, pp. 813-826, 1987.
[24] A. Ravichandran, R. Khare, and C. C. Chen, "Predicting NRTL Binary Interaction
Parameters from Molecular Simulations," AIChE Journal, 2018.
[25] B. Chen, J. I. Siepmann, and M. L. Klein, "Direct Gibbs Ensemble Monte Carlo
Simulations for Solid−Vapor Phase Equilibria:  Applications to
Lennard−Jonesium and Carbon Dioxide," The Journal of Physical Chemistry B,
vol. 105, pp. 9840-9848, 2001.
[26] E. Sanz and C. Vega, "Solubility of KF and NaCl in Water by Molecular
Simulation," The Journal of Chemical Physics, vol. 126, p. 014507, 2007.
[27] Z. Mester and A. Z. Panagiotopoulos, "Temperature-Dependent Solubilities and
Mean Ionic Activity Coefficients of Alkali Halides in Water From Molecular
Dynamics Simulations," The Journal of Chemical Physics, vol. 143, p. 044505,
2015.
[28] Z. Mester and A. Z. Panagiotopoulos, "Mean Ionic Activity Coefficients in
Aqueous NaCl Solutions from Molecular Dynamics Simulations," The Journal of
Chemical Physics, vol. 142, p. 044507, 2015.
[29] J. G. Kirkwood and F. P. Buff, "The Statistical Mechanical Theory of Solutions.
I," The Journal of Chemical Physics, vol. 19, pp. 774-777, 1951.
[30] S. Weerasinghe and P. E. Smith, "A Kirkwood–Buff Derived Force Field for
Sodium Chloride in Water," The Journal of Chemical Physics, vol. 119, pp.
11342-11349, 2003.

41
Texas Tech University, Nazir Hossain, August 2018

[31] M. B. Gee, N. R. Cox, Y. Jiao, N. Bentenitis, S. Weerasinghe, and P. E. Smith, "A


Kirkwood-Buff Derived Force Field for Aqueous Alkali Halides," Journal of
Chemical Theory and Computation, vol. 7, pp. 1369-1380, 2011.
[32] S. Weerasinghe and P. E. Smith, "Kirkwood–Buff Derived Force Field for
Mixtures of Acetone and Water," The Journal of Chemical Physics, vol. 118, pp.
10663-10670, 2003.
[33] H. Jiang, Z. Mester, O. A. Moultos, I. G. Economou, and A. Z. Panagiotopoulos,
"Thermodynamic and Transport Properties of H2O + NaCl from Polarizable Force
Fields," Journal of Chemical Theory and Computation, vol. 11, pp. 3802-3810,
2015.
[34] R. A. Robinson and R. H. Stokes, Electrolyte Solutions: Courier Dover
Publications, 2002.
[35] W. J. Hamer and Y. C. Wu, "Osmotic Coefficients and Mean Activity
Coefficients of Uni‐univalent Electrolytes in Water at 25 °C," Journal of Physical
and Chemical Reference Data, vol. 1, pp. 1047-1100, 1972.
[36] I. N. Tang, H. R. Munkelwitz, and N. Wang, "Water Activity Measurements with
Single Suspended Droplets: The NaCl-H2O and KCl-H2O Systems," Journal of
Colloid and Interface Science, vol. 114, pp. 409-415, 1986.
[37] C. K. Chan, Z. Liang, J. Zheng, S. L. Clegg, and P. Brimblecombe,
"Thermodynamic Properties of Aqueous Aerosols to High Supersaturation: I—
Measurements of Water Activity of the System Na+−Cl−−NO3−−SO24−−H2O at ~
298.15 K," Aerosol Science and Technology, vol. 27, pp. 324-344, 1997.
[38] A. Ben-Naim, Molecular theory of solutions: Oxford University Press, 2006.
[39] R. Cortes-Huerto, K. Kremer, and R. Potestio, "Communication: Kirkwood-Buff
Integrals in the Thermodynamic Limit from Small-Sized Molecular Dynamics
Simulations," The Journal of Chemical Physics, vol. 145, p. 141103, 2016.
[40] P. Ganguly and N. F. A. van der Vegt, "Convergence of Sampling Kirkwood–
Buff Integrals of Aqueous Solutions with Molecular Dynamics Simulations,"
Journal of Chemical Theory and Computation, vol. 9, pp. 1347-1355, 2013.
[41] H. J. C. Berendsen, J. R. Grigera, and T. P. Straatsma, "The Missing Term in
Effective Pair Potentials," The Journal of Physical Chemistry, vol. 91, pp. 6269-
6271, 1987.
[42] S. Plimpton, "Fast Parallel Algorithms for Short-Range Molecular Dynamics,"
Journal of Computational Physics, vol. 117, pp. 1-19, 1995.
[43] W. Shinoda, M. Shiga, and M. Mikami, "Rapid Estimation of Elastic Constants
By Molecular Dynamics Simulation Under Constant Stress," Physical Review B,
vol. 69, p. 134103, 2004.
[44] R. W. Hockney and J. W. Eastwood, Computer Simulation using Particles: crc
Press, 1988.
[45] W. H. Press, S. A. Teukolsky, W. T. Vetterling, and B. P. Flannery, Numerical
Recipes in C vol. 2: Cambridge University Press Cambridge, 1996.
[46] ASPEN Physical Property System, V 8.4, ASPEN Technology Inc., , Burlington,
MA, 2013.

42
Texas Tech University, Nazir Hossain, August 2018

[47] Y. Luo, W. Jiang, H. Yu, A. D. MacKerell, and B. Roux, "Simulation Study of


Ion Pairing in Concentrated Aqueous Salt Solutions with A Polarizable Force
Field," Faraday Discussions, vol. 160, pp. 135-149, 2013.

43
Texas Tech University, Nazir Hossain, August 2018

CHAPTER 3
TEMPERATURE DEPENDENCE OF INTERACTION
PARAMETERS IN ELECTROLYTE NRTL MODEL
3.1 Abstract
The electrolyte NRTL model captures the electrolyte solution nonideality over the

entire concentration range with two binary interaction parameters. Here the temperature

dependence of the binary parameters is formulated with a Gibbs-Helmholtz expression

containing three temperature coefficients associated with Gibbs energy, enthalpy, and

heat capacity contributions. We show the Gibbs energy term is correlated to the excess

Gibbs energy of aqueous single electrolyte systems at 298.15 K. With the Gibbs energy

term identified, the enthalpy term is correlated to the excess enthalpy at 298.15 K. With

the Gibbs energy term and the enthalpy term identified, the heat capacity term is

correlated to the excess heat capacity at 298.15 K. Once the temperature coefficients are

properly quantified by regressing data of mean ionic activity coefficient, excess enthalpy,

and heat capacity at 298.15 K or the equivalents, the model provides a comprehensive

thermodynamic framework to represent all thermodynamic properties of electrolyte

solutions.

Keywords: Electrolyte NRTL model; binary interaction parameters; temperature

dependence; excess Gibbs energy, excess enthalpy; heat capacity

44
Texas Tech University, Nazir Hossain, August 2018

3.2 Introduction
Thermodynamic modeling of aqueous electrolyte solutions is an important and

challenging task [1]. For example, rapidly increased hydraulic fracturing activities in oil

and gas industry generate enormous amounts of “produced water” that causes public

health and environmental concerns [2]. Common constituents of produced water are

dissolved and dispersed oil, dissolved formation minerals including electrolytes, heavy

metals and radioactive materials, production chemical compounds, production solids such

as formation solids, corrosion and scale formation, waxes and asphaltenes, and dissolved

gases. The presence of cations such as Na+, K+, Ca2+, Mg2+, Ba2+, Sr2+, Fe2+ and their

counter anions such as Cl−, SO42−, CO32−, HCO3− is responsible for the complex

chemistry of produced water in terms of buffering capacity, salinity and scale formation

[3]. Development of desalination or treatment processes such as multi-stage flash (MSF)

and multi-effect distillation (MED) for such high salinity brine requires accurate

calculations of vapor pressure, calorimetric properties, and salt solubility [4]. The Pitzer

model [5], the extended UNIQUAC (EUNIQUAC) [6, 7] model, the OLI-MSE model [8,

9] and the electrolyte non-random two liquid (eNRTL) theory [10-12] are among the

most commonly used thermodynamic models for aqueous electrolyte systems.

The Pitzer model has been widely used for modeling electrolyte solutions in recent

decades. It represents the excess Gibbs energy as a virial expansion of two body and three

body interactions. Although the model has a theoretical basis, its formulation and

parameters are purely empirical functions of ionic strength and temperature. For example,

(0) (1) 𝛾
the Pitzer model parameters, 𝛽𝑀𝑋 , 𝛽𝑀𝑋 , and 𝐶𝑀𝑋 , have the following polynomial

45
Texas Tech University, Nazir Hossain, August 2018

function form for aqueous NaCl solution with respect to temperature (T) and pressure (P)

[13].
𝑎1 𝑎9 +𝑎10 𝑃
𝑓(𝑇) = + 𝑎2 + 𝑎3 𝑃 + 𝑎4 𝑙𝑛(𝑇) + (𝑎5 + 𝑎6 𝑃)𝑇 + (𝑎7 + 𝑎8 𝑃)𝑇 2 + +
𝑇 𝑇−227

𝑎11 +𝑎12 𝑃
(3.1)
680−𝑇

here 𝑎𝑖 ′𝑠 are temperature and pressure coefficients. Furthermore, the functional form and

the number of adjustable temperature and pressure coefficients vary with electrolytes

(0) (1)
[13]. Figure 3.1 shows the highly nonlinear nature of the model parameters, 𝛽𝑀𝑋 , 𝛽𝑀𝑋 ,
𝛾
and 𝐶𝑀𝑋 , for NaCl + H2O and Na2SO4 + H2O binary systems with respect to temperature.

Moreover, due to the inherent limitation of virial expansion equations, the Pitzer model

does not extrapolate reliably for thermodynamic properties of electrolyte solutions

beyond ionic strength of 6 to 10 molal [14]. Use of the Pitzer model in heat and mass

balance calculations and process simulation applications for electrolyte solutions proves

to be extremely cumbersome if not impossible.

Revised Extended UNIQUAC model [7], named as REUNIQUAC, expresses the excess

Gibbs energy equation as a combination of the Debye-Hückel term and the UNIQUAC

term. The UNIQAUC terms consists of combinatorial and residual contributions

accounting for size effects and interaction effects respectively. In the residual

contribution, the interaction parameters (𝑢𝑗𝑖 ′𝑠) are expressed as a quadratic function of

temperature [7]:

(0) (1) (2)


𝑢𝑗𝑖 (𝑇) = 𝑢𝑗𝑖 + 𝑢𝑗𝑖 𝑇 + 𝑢𝑗𝑖 𝑇 2 (3.2)

46
Texas Tech University, Nazir Hossain, August 2018

(0) (1) (2)


Where 𝑢𝑗𝑖 , 𝑢𝑗𝑖 , and 𝑢𝑗𝑖 are temperature coefficients. The REQUNIQUAC model is

only applicable up to 5 molal [7].

OLI-MSE model expresses the excess Gibbs energy of aqueous electrolyte

solutions in terms of long range, middle range, and short range interactions [9]. The long

range contribution is calculated from Debye-Hückel limiting law and the short range

contribution is calculated from UNIQUAC model. The model further introduces a virial

equation with second virial coefficients for the “middle range” ion-ion interaction

contribution. The “middle range” interaction parameter, 𝐵𝑖𝑗 , is an empirical function of

ionic strength (𝐼𝑥 ) and temperature expressed as:

𝐵𝑖𝑗 (𝐼𝑥 ) = 𝑏𝑖𝑗 + 𝑐𝑖𝑗 𝑒𝑥𝑝 (−√(𝐼𝑥 + 0.01)) (3.3)

𝑏2,𝑖𝑗
𝑏𝑖𝑗 = 𝑏0,𝑖𝑗 + 𝑏1,𝑖𝑗 𝑇 + ⁄ (3.4)
𝑇
𝑐2,𝑖𝑗
𝑐𝑖𝑗 = 𝑐0,𝑖𝑗 + 𝑐1,𝑖𝑗 𝑇 + ⁄𝑇 (3.5)

The electrolyte NRTL [10] model represents excess Gibbs energy of electrolyte

solutions as the sum of two contributions: long range and short range. The long range

electrostatic contribution is calculated from the Pitzer-Debye-Hückel limiting law

equation. For the short range contribution, the non-random two-liquid theory is

considered together with two hypotheses for the liquid lattice structure: 1) like ion

repulsion and 2) local electroneutrality. Two adjustable binary interaction parameters are

required to model aqueous single electrolyte systems and they are found to be weak and

largely linear function of temperature. A Gibbs-Helmholtz like equation suggested by

47
Texas Tech University, Nazir Hossain, August 2018

Clarke and Glew [15] was proposed for the temperature dependence of the binary

interaction parameters:

1 1 𝑇𝑟𝑒𝑓 −𝑇 𝑇
𝜏𝑖𝑗 = 𝑎𝑖𝑗 + 𝑏𝑖𝑗 (𝑇 − 𝑇 ) + 𝑐𝑖𝑗 ( + 𝑙𝑛 𝑇 ) (3.6)
𝑟𝑒𝑓 𝑇 𝑟𝑒𝑓

As an example, the eNRTL binary interaction parameters for NaCl + H2O and

Na2SO4 + H2O binary systems have been reported by Yan and Chen [16] by

simultaneously regressing experimental data on mean ionic activity coefficient, osmotic

coefficient, liquid enthalpy, and heat capacity of electrolyte solutions with electrolyte

concentrations up to saturation and temperatures up to 473.15 K. Three temperature

coefficients are sufficient to describe the relatively linear temperature dependence of the

binary interaction parameters. The temperature dependence of the eNRTL binary

interaction parameters for these two systems are shown in Figure 3.2.

In this work, we present a systematic examination on the temperature dependence

of the eNRTL binary interaction parameters in terms of the Gibbs-Helmholtz equation

with specific contribution terms from Gibbs energy, enthalpy, and heat capacity. We

show the Gibbs energy term of the interaction parameters is correlated to the excess

Gibbs energy of binary systems at 298.15 K. With the Gibbs energy term identified, the

enthalpy term is correlated to the excess enthalpy at 298.15 K. With the Gibbs energy

term and the enthalpy term identified, the heat capacity term is correlated to the excess

heat capacity at 298.15 K. The resulting eNRTL model and the parameters represent all

thermodynamic properties of aqueous electrolyte systems such as vapor pressure, mean

ionic activity coefficients, osmotic coefficients, molar excess enthalpy, heat capacity, etc.

48
Texas Tech University, Nazir Hossain, August 2018

over the entire ranges of concentration and temperature. ASPEN 8.4 is used for

regression and data analysis.

3.3 Thermodynamic Framework


The eNRTL model accounts for the long range and short range interaction

contributions to the excess Gibbs energy as follows:

𝐺 𝑒𝑥 = 𝐺 𝑒𝑥,𝑃𝐷𝐻 + 𝐺 𝑒𝑥,𝑙𝑐 (3.7)

where 𝐺 𝑒𝑥 is the total excess Gibbs energy, 𝐺 𝑒𝑥,𝑃𝐷𝐻 is the contribution from long range

ion-ion interactions, and 𝐺 𝑒𝑥,𝑙𝑐 is the excess Gibbs energy contribution from short range

(local) ion-ion, ion-molecule and molecule-molecule interactions.

The long range ion-ion interaction contribution, the PDH term, is calculated from

the Pitzer-Debye-Hückel (PDH) formula as given below


1/2
𝐺 𝑒𝑥,𝑃𝐷𝐻 4𝐴𝜙 𝐼𝑥 1+𝜌𝐼𝑥
=− 𝑙𝑛 [ 1/2 ] (3.8)
𝑛𝑅𝑇 𝜌 1+𝜌(𝐼𝑥0 )

3/2
1 2𝜋𝑁𝐴 1/2 𝑄𝑒2
𝐴𝜙 = 3 ( ) ( ) (3.9)
𝑣 𝜖𝑘𝐵 𝑇

1 1 1
𝐼𝑥 = 2 ∑𝑖 𝑧𝑖2 𝑥𝑖 = 2 ∑𝑐 𝑧𝑐2 𝑥𝑐 + 2 ∑𝑎 𝑧𝑎2 𝑥𝑎 (3.10)

1 1
𝜖 = 𝐴 + 𝐵 (𝑇 − 𝐶 ) (3.11)

where 𝐴𝜙 is the Debye-Hückel parameter; 𝐼𝑥 is the ionic strength in mole fraction; 𝜌 is

the closest approach parameter; 𝐼𝑥0 is the ionic strength at fused salt reference state; 𝑁𝐴 is

Avogadro's number; 𝑄𝑒 is the electron charge, 4.80298×10-10 E.S.U; 𝑘𝐵 is Boltzmann

constant, 1.38054×10-16; v and 𝜀 are the molar volume, in cm3/mol and the dielectric

constant of water, respectively; 𝑧𝑖 is the charge number of species i; 𝑧𝑐 and 𝑧𝑎 are the

charge numbers of cation and anion, respectively. 𝜀 is correlated with temperature as


49
Texas Tech University, Nazir Hossain, August 2018

shown in Eq. 3.11, where A, B, C are the correlation parameters. The values of A, B and

C for solvent water are 78.54, 31989.4 and 298.15, respectively.

The short range interaction contribution, the lc term, is calculated from the local

composition model of the non-random two-liquid theory:

𝐺 𝑒𝑥,𝑙𝑐 ∑𝑖 𝑋𝑖 𝐺𝑖𝑚 𝜏𝑖𝑚 ∑𝑖≠𝑐 𝑋𝑖 𝐺𝑖𝑐 𝜏𝑖𝑐 ∑𝑖≠𝑎 𝑋𝑖 𝐺𝑖𝑎 𝜏𝑖𝑎


= ∑𝑚 𝑛𝑚 ( ∑𝑖 𝑋𝑖 𝐺𝑖𝑚
) + ∑𝑐 𝑧𝑐 𝑛𝑐 ( ∑𝑖≠𝑐 𝑋𝑖 𝐺𝑖𝑐
) + ∑𝑎 𝑧𝑎 𝑛𝑎 ( ∑𝑖≠𝑎 𝑋𝑖 𝐺𝑖𝑎
) (3.12)
𝑛𝑅𝑇

𝑛
𝑋𝑖 = 𝐶𝑖 𝑥𝑖 = 𝐶𝑖 ( 𝑛𝑖 ), i = m,c,a (3.13)

𝑛 = ∑𝑖 𝑛𝑖 = ∑𝑚 𝑛𝑚 + ∑𝑐 𝑛𝑐 + ∑𝑎 𝑛𝑎 (3.14)

𝐺𝑖𝑗 = 𝑒𝑥𝑝(−𝛼𝑖𝑗 𝜏𝑖𝑗 ) (3.15)

where the indices m, c and a correspond to molecular, cationic and anionic species

respectively; 𝑛𝑖 and 𝑥𝑖 are the mole number and the mole fraction of species i in the

system, respectively; 𝐶𝑖 is the absolute charge number for ionic species and unity for

molecular species; αij is the non-randomness factor (fixed at 0.2 in this study); 𝜏𝑖𝑗 is the

asymmetric binary interaction energy parameter. Note that, for single electrolyte systems

with solvent water (w) and electrolyte (ca), 𝜏𝑖𝑗 ′𝑠 are condensed and expressed in terms of

water-electrolyte binary interaction parameters, 𝜏𝑤,𝑐𝑎 and electrolyte-water binary

interaction parameters, 𝜏𝑐𝑎,𝑤 .

The binary interaction parameter, 𝜏𝑖𝑗 , is a weak and well-behaved function of

temperature. The temperature dependence can be written with the Gibbs-Helmholtz type

equation as [15]:

1 1 𝑇𝑟𝑒𝑓 −𝑇 𝑇
𝜏𝑖𝑗 = 𝛥𝑔𝑖𝑗 + 𝛥ℎ𝑖𝑗 (𝑇 − 𝑇 ) + ∆𝑐𝑝 𝑖𝑗 ( + 𝑙𝑛 𝑇 ) (3.16)
𝑟𝑒𝑓 𝑇 𝑟𝑒𝑓

50
Texas Tech University, Nazir Hossain, August 2018

where ∆𝑔𝑖𝑗 , ∆ℎ𝑖𝑗 , and ∆𝑐𝑝 𝑖𝑗 are Gibbs energy, enthalpy, and heat capacity contributions

to the binary interaction parameters and 𝑇𝑟𝑒𝑓 is a reference temperature fixed at 298.15 K.

The activity coefficient, 𝛾𝑖 , of component i can be expressed as follows:

1 𝜕𝐺 𝑒𝑥,𝑃𝐷𝐻 1 𝜕𝐺 𝑒𝑥,𝑙𝑐
𝑙𝑛𝛾𝑖 = 𝑙𝑛𝛾𝑖𝑃𝐷𝐻 + 𝑙𝑛𝛾𝑖𝑙𝑐 = 𝑅𝑇 ( ) + 𝑅𝑇 ( ) , i= c, a, w (3.17)
𝜕𝑛𝑖 𝑇,𝑃,𝑛𝑗≠𝑖 𝜕𝑛𝑖 𝑇,𝑃,𝑛𝑗≠𝑖

The long range contributions to the activity coefficients for molecular and ionic species

are given by following equations


3/2
𝑃𝐷𝐻 2𝐴𝜙 𝐼𝑥
𝑙𝑛𝛾𝑚 = 1/2 (3.18)
1+𝜌𝐼𝑥

1 1 3
−1/2
2𝑧𝑖2 1+𝜌𝐼𝑥2 𝑧𝑖2 𝐼𝑥2 −2𝐼𝑥2 2𝐼𝑥 (𝐼𝑥0 ) 𝜕𝐼 0
𝑙𝑛𝛾𝑖𝑃𝐷𝐻 = −𝐴𝜙 (( ) 𝑙𝑛 ( 1 )+ 1 + 1 (𝑛 𝜕𝑛𝑥 )) , i = c,a (3.19)
𝜌 𝑖
1+𝜌(𝐼𝑥0 )2 1+𝜌𝐼𝑥2 1+𝜌(𝐼𝑥0 )2

For aqueous phase infinite dilution reference state:

𝜕𝐼𝑥0
𝐼𝑥0 = 0 and = 0 , i = c,a (3.20)
𝜕𝑛𝑖

The corresponding short range contribution to the activity coefficients of species

in aqueous single electrolyte solutions is:


𝑋𝑐 𝜏𝑐𝑤 𝐺𝑐𝑤 +𝑋𝑎 𝜏𝑎𝑤 𝐺𝑎𝑤 𝑋 𝑋 𝜏 𝐺𝑤𝑐 𝑋 𝑋 𝜏 𝐺𝑤𝑎 𝑋𝑐 𝑋𝑤 𝜏𝑐𝑤 𝐺𝑐𝑤 +𝑋𝑎 𝑋𝑤 𝜏𝑎𝑤 𝐺𝑎𝑤
𝑙𝑛𝛾𝑤𝑙𝑐 = (𝑋𝑐 𝐺𝑐𝑤 +𝑋𝑎 𝐺𝑎𝑤 +𝑋𝑤 )
+ (𝑋𝑐 +𝑋
𝑎 𝑤𝑐
2 + (𝑋𝑎 +𝑋
𝑐 𝑤𝑎
2 − (𝑋𝑐 𝐺𝑐𝑤 +𝑋𝑎 𝐺𝑎𝑤 +𝑋𝑤 )2
(3.21)
𝑎 𝐺 𝑤 𝑤𝑐 ) 𝑐 𝐺 𝑤 𝑤𝑎 )

1 2𝜏 𝐺
𝑋𝑤 𝑋 𝑋 𝜏 𝐺𝑤𝑎 𝑋 𝜏 𝐺
𝑙𝑛𝛾𝑐𝑙𝑐 = (𝑋 𝑐𝑤 𝑐𝑤
2
+ (𝑋𝑎 +𝑋
𝑤 𝑤𝑎
2
+ (𝑋 𝑤+𝑋𝑤𝑐 𝐺𝑤𝑐 ) (3.22)
𝑍𝑐 𝑐 𝐺𝑐𝑤 +𝑋𝑎 𝐺𝑎𝑤 +𝑋𝑤 ) 𝑐 𝐺 𝑤 𝑤𝑎 ) 𝑎 𝑤 𝑤𝑐

1 2𝜏
𝑋𝑤 𝑎𝑤 𝐺𝑎𝑤 𝑋 𝑋 𝜏 𝐺𝑤𝑐 𝑋 𝜏 𝐺
𝑙𝑛𝛾𝑎𝑙𝑐 = (𝑋 2
+ (𝑋𝑐 +𝑋
𝑤 𝑤𝑐
2
+ (𝑋𝑤+𝑋𝑤𝑎 𝐺𝑤𝑎) (3.23)
𝑍𝑎 𝑐 𝐺𝑐𝑤 +𝑋𝑎 𝐺𝑎𝑤 +𝑋𝑤 ) 𝑎 𝐺 𝑤 𝑤𝑐 ) 𝑐 𝑤 𝑤𝑐

The unsymmetric activity coefficients of ionic species with aqueous phase infinite

dilution reference state are:

𝑙𝑛𝛾𝑖∗𝑙𝑐 = 𝑙𝑛𝛾𝑖𝑙𝑐 − 𝑙𝑛𝛾𝑖𝑙𝑐,∞ , i = c, a (3.24)

51
Texas Tech University, Nazir Hossain, August 2018

where,

𝑙𝑛𝛾𝑖𝑙𝑐,∞ = 𝜏𝑤𝑖 + 𝐺𝑖𝑤 𝜏𝑖𝑤 , i = c, a (3.25)

Consider 1 mole of electrolyte ca dissociates as follows:

𝑐𝑎 ↔ 𝜈𝑐 𝑐 +𝑧𝑐 + 𝜈𝑎 𝑐 −𝑧𝑎 (3.26)

where 𝜈𝑐 and 𝜈𝑎 are the stoichiometric coefficients for cation and anion respectively. The

mole fraction scale mean ionic activity coefficient (𝛾±∗) is given by following equation:
1
𝛾±∗ = (𝛾𝑐∗ 𝜈𝑐 𝛾𝑎∗ 𝜈𝑎 )(𝜈𝑐 +𝜈𝑎) (3.27)


The molality scale mean ionic activity coefficient (𝛾±𝑚 ) is defined as:

∗ (𝜈𝑐 +𝜈𝑎 )𝑚𝑀𝑠


𝑙𝑛𝛾±𝑚 = 𝑙𝑛𝛾±∗ − 𝑙 𝑛 (1 + ) (3.28)
1000

where 𝑚 is the molality of the electrolyte in solution and 𝑀𝑠 is the molecular weight of

solvent.

Excess molar enthalpy (ℎ∗𝑒𝑥 ) is calculated from temperature derivatives of logarithm of

activity coefficients:

𝜕𝑙𝑛𝛾𝑖 𝜕
ℎ∗𝑒𝑥 = −𝑅𝑇 2 ∑𝑖 𝑥𝑖 = −𝑅𝑇 2 ∑𝑖 𝑥𝑖 𝜕𝑇 (𝑙𝑛𝛾𝑖𝑃𝐷𝐻 + 𝑙𝑛𝛾𝑖𝑙𝑐 ), i = c, a, w (3.29)
𝜕𝑇

Since the excess enthalpy, ℎ∗𝑒𝑥 , is based on aqueous phase infinite dilution reference

state for ions the activity coefficient of ions in Eq. 29 are unsymmetric ones with aqueous

phase infinite dilution reference state whereas the activity coefficient of water is

symmetric.

In the long range contribution to the activity coefficients, 𝑙𝑛𝛾𝑖𝑃𝐷𝐻 , only 𝐴𝜙 is

temperature dependent and there is no adjustable parameter. We calculate the PDH

contribution to the excess enthalpy numerically through ASPEN. In the short range

52
Texas Tech University, Nazir Hossain, August 2018

contribution to the activity coefficients, 𝑙𝑛𝛾𝑖𝑙𝑐 , 𝜏𝑖𝑗 and 𝐺𝑖𝑗 are temperature dependent.

Following Eq. 16, the short range contribution to the excess enthalpy is determined by

𝜏𝑖𝑗 , 𝐺𝑖𝑗 and their temperature derivatives as shown below:

𝜕𝜏𝑖𝑗 𝛥ℎ𝑖𝑗 1 𝑇𝑟𝑒𝑓


=− + ∆𝑐𝑝 𝑖𝑗 (𝑇 − ) (3.30)
𝜕𝑇 𝑇2 𝑇2

𝜕𝐺𝑖𝑗 𝛥ℎ𝑖𝑗 1 𝑇𝑟𝑒𝑓


= (−𝛼)𝑒𝑥𝑝(−𝛼𝜏𝑖𝑗 ) (− + ∆𝑐𝑝 𝑖𝑗 (𝑇 − )) (3.31)
𝜕𝑇 𝑇2 𝑇2

𝜕𝜏𝑖𝑗 𝜕𝐺𝑖𝑗
At 𝑇 = 𝑇𝑟𝑒𝑓 , the ∆𝑐𝑝 𝑖𝑗 terms cancel out and they have no contribution to 𝜏𝑖𝑗 , 𝐺𝑖𝑗 , ,
𝜕𝑇 𝜕𝑇

and ℎ∗𝑒𝑥 . In other words, the short range contribution to the excess enthalpy at 298.15 K

is determined by the first two temperature coefficients of the eNRTL binary interaction

parameters, i.e., ∆𝑔𝑖𝑗 and ∆ℎ𝑖𝑗 .

Liquid molar enthalpy, ℎ𝑙 , is expressed as follows:

ℎ𝑙 = 𝑥𝑤 ℎ𝑤
0
+ ∑𝑘 𝑥𝑘 ℎ𝑘∞,𝑎𝑞 + ℎ∗𝑒𝑥 (3.32)

𝑜 𝑖𝑔 𝑇 𝑖𝑔 𝑖𝑔
ℎ𝑤 = ∆𝑓 ℎ𝑤 + ∫298.15 𝑐𝑝,𝑤 𝑑𝑇 + (ℎ𝑤 (𝑇, 𝑃) − ℎ𝑤 (𝑇, 𝑃)) (3.33)

𝑇
ℎ𝑘∞,𝑎𝑞 = ∆𝑓 ℎ𝑘∞,𝑎𝑞 + ∫298.15 𝑐𝑝,𝑘
∞,𝑎𝑞
𝑑𝑇 ; 𝑘 = 𝑐, 𝑎 (3.34)

0 𝑖𝑔
where ℎ𝑤 is the liquid molar enthalpy of water; ∆𝑓 ℎ𝑤 is the ideal gas enthalpy of

𝑖𝑔
formation of water; 𝑐𝑝,𝑤 is the ideal gas specific heat capacity of water; (ℎ𝑤 (𝑇, 𝑃) −

𝑖𝑔
ℎ𝑤 (𝑇, 𝑃)) is the liquid water molar enthalpy departure from ideal gas; ℎ𝑘∞,𝑎𝑞 is the molar

enthalpy of ionic species k at the aqueous phase infinite dilution reference state; ∆𝑓 ℎ𝑘∞,𝑎𝑞

∞,𝑎𝑞
the enthalpy of formation; 𝑐𝑝,𝑘 is the specific heat capacity.

The liquid molar heat capacity, 𝑐𝑝 and the excess liquid molar heat capacity, 𝑐𝑝∗𝑒𝑥 , can be

calculated by following equations:


53
Texas Tech University, Nazir Hossain, August 2018

𝜕ℎ𝑙
𝑐𝑝 = ( 𝜕𝑇 ) (3.35)
𝑝

∞,𝑎𝑞
𝑐𝑝∗𝑒𝑥 = 𝑐𝑝 − 𝑥𝑤 𝑐𝑝,𝑤 − ∑𝑘 𝑥𝑘 𝑐𝑝,𝑘 (3.36)

The calculations of 𝑐𝑝 and 𝑐𝑝∗𝑒𝑥 require second temperature derivatives of the

activity coefficients. The long range contribution has no adjustable parameters and we

calculate the PDH contribution to the excess heat capacity numerically through ASPEN.

In the short range term, the second temperature derivatives of 𝜏𝑖𝑗 and 𝐺𝑖𝑗 involve the

following equations:

𝜕2 𝜏𝑖𝑗 2𝛥ℎ𝑖𝑗 1 2𝑇𝑟𝑒𝑓


= + ∆𝑐𝑝 𝑖𝑗 (− 𝑇 2 + ) (3.37)
𝜕𝑇 2 𝑇3 𝑇3

𝜕2 𝐺𝑖𝑗 𝜕2 𝜏 𝜕𝜏 2
= (−𝛼)𝑒𝑥𝑝(−𝛼𝜏𝑖𝑗 ) ( 𝜕𝑇 2𝑖𝑗 − ( 𝜕𝑇𝑖𝑗) ) (3.38)
𝜕𝑇 2

Therefore heat capacity of aqueous electrolyte solutions at 298.15 K depends on 𝑐𝑝,𝑤 ,

∞,𝑎𝑞
𝑐𝑝,𝑘 and the temperature coefficients of the eNRTL binary interaction parameters, i.e.

𝛥𝑔𝑖𝑗 , 𝛥ℎ𝑖𝑗 , and ∆𝑐𝑝 𝑖𝑗 .

Therefore, at 𝑇 = 𝑇𝑟𝑒𝑓 , the excess heat capacity of aqueous electrolyte solutions depends

on the three temperature coefficients of the eNRTL binary interaction parameters, i.e.

𝛥𝑔𝑖𝑗 , 𝛥ℎ𝑖𝑗 , and ∆𝑐𝑝 𝑖𝑗 .

Note that all equations above for molar enthalpy and molar heat capacity are

written with respect to true species. Proper conversion to apparent components is

necessary if apparent electrolyte concentrations are reported.

54
Texas Tech University, Nazir Hossain, August 2018

3.4 Temperature Dependence of Interaction Parameters


The advantage of expressing the temperature dependence of the eNRTL binary

interaction parameters with Eq. 3.16 is that the three temperature coefficients are

separable and can be individually quantified by regression with pertinent thermodynamic

property data. For example, the Gibbs energy terms, 𝛥𝑔𝑤,𝑐𝑎 and 𝛥𝑔𝑐𝑎,𝑤 , of the binary

interaction parameters for the water-electrolyte (cation c, anion a) pair can be regressed

from mean ionic activity coefficient, osmotic coefficient, or vapor pressure depression

data at 298.15 K since the enthalpy term and the heat capacity term have no effect on

excess Gibbs energy at 298.15 K. To limit the scope of this work, we focus on mean ionic

activity coefficient data for regression. Osmotic coefficient data, vapor pressure

depression data and mean ionic activity coefficient data are equivalent among them

because they are related to each other via Gibbs-Duhem equation for completely

dissociated aqueous single electrolytes. Figure 3.3 shows the family of curves for

experimental mole fraction scale mean ionic activity coefficients of 41 1-1 electrolytes

vs. electrolyte mole fraction [17]. The mean ionic activity coefficients of the electrolytes

at 298.15 K are well correlated by 𝛥𝑔𝑤,𝑐𝑎 and 𝛥𝑔𝑐𝑎,𝑤 . Table 1 shows the regressed values

for 𝛥𝑔𝑤,𝑐𝑎 and 𝛥𝑔𝑐𝑎,𝑤 for 70 1-1, 1-2, 2-1 and 2-2 electrolytes. In regression, the

standard deviation for the mean ionic activity coefficient is set to 2%. Interestingly, the

values of 𝛥𝑔𝑤,𝑐𝑎 for all 70 electrolytes are positive and fall in the range of 7 to 12 while

the values of 𝛥𝑔𝑐𝑎,𝑤 are all negative. The ratios of |𝛥𝑔𝑤,𝑐𝑎 | and |𝛥𝑔𝑐𝑎,𝑤 | for the aqueous

electrolytes are in the range of 1.8 to 2.2.

Figure 3.4 shows the mean ionic activity coefficients of 1-1 electrolytes at 298.15

K calculated with varying values for (𝛥𝑔𝑤,𝑐𝑎 , 𝛥𝑔𝑐𝑎,𝑤 ) with the ratio of |𝛥𝑔𝑤,𝑐𝑎 | and
55
Texas Tech University, Nazir Hossain, August 2018

|𝛥𝑔𝑐𝑎,𝑤 | fixed at 2. Also shown in Figure 4 are the mean ionic activity coefficients

calculated by the PDH term. Larger values of 𝛥𝑔𝑤,𝑐𝑎 yield larger mean ionic activity

coefficients. With (𝛥𝑔𝑤,𝑐𝑎 , 𝛥𝑔𝑐𝑎,𝑤 ) = (6.932, −3.466), the calculated mean ionic activity

coefficients overlap with the values calculated by the PDH term alone. In other words, at

these values for ∆𝑔𝑖𝑗 , the short range contributions to the activity coefficients are zero.

Figure 3.5 presents the short range contribution, the lc term, to the symmetric

excess Gibbs energy of aqueous 1-1 electrolytes at 298.15 K calculated with varying

values for (𝛥𝑔𝑤,𝑐𝑎 , 𝛥𝑔𝑐𝑎,𝑤 ) and |𝛥𝑔𝑤,𝑐𝑎 |/|𝛥𝑔𝑐𝑎,𝑤 |=2. Also shown is the PDH

contribution which is found to be positive. With (𝛥𝑔𝑤,𝑐𝑎 , 𝛥𝑔𝑐𝑎,𝑤 ) = (6.932, −3.466), the

lc term contribution to symmetric excess Gibbs energy is zero. With 𝛥𝑔𝑤,𝑐𝑎 < 6.932, the

lc term contribution is positive. Larger values of 𝛥𝑔𝑤,𝑐𝑎 yield more negative excess

Gibbs energy.

Figure 3.6 presents symmetric excess Gibbs energy for some representative

aqueous electrolytes at 298.15 K calculated with the regressed binary parameters shown

in Table 1. More negative excess Gibbs energy suggests stronger electrolyte-water

interaction, more stable liquid lattice structure, and therefore “structure making”. On the

contrary, positive excess Gibbs energy suggests weaker electrolyte-water interaction, less

stable liquid lattice structure, and “structure breaking” [18].

Excess enthalpy corresponds to the deviation of enthalpy from ideal solution.

Wagman et al. [19] compiled data on liquid molar enthalpy of aqueous single electrolytes

at 298.15 K with aqueous phase infinite dilution reference state. From the enthalpy data

of Wagman et al. the excess molar enthalpy of aqueous single electrolyte systems can be

56
Texas Tech University, Nazir Hossain, August 2018

calculated. Figure 3.7 presents the calculated excess molar enthalpies at 298.15 K for 42

1-1 aqueous single electrolytes as a function of electrolyte concentration. The family of

curves shows the excess molar enthalpies are largely linear with respect to electrolyte

concentration with both positive and negative slopes. However, some of the electrolytes

such as strong acids (HCl, HBr, HI, HClO4, HNO3, etc.) and caustic (KOH and NaOH)

show stiff slopes compared to others. The observed positive and upward shifted slopes of

the excess enthalpy are the result of incomplete dissociation of the electrolytes and

hydration of the ionic species [20]. These incompletely dissociated electrolytes are

outside the scope of this study since exact account of speciation would be required.

The excess enthalpy is related to the temperature derivative of the activity

coefficients as shown in Eq. 3.29. The stronger (positive) the temperature dependence,

the more negative the excess enthalpy becomes. The excess enthalpy due to the long

range ion-ion interaction, i.e., the PDH term, varies with each class of aqueous

electrolytes. Figure 3.8 shows the calculated excess enthalpy contribution from the PDH

term for aqueous 1-1, 1-2 (2-1), 1-3 (3-1) and 2-2 single electrolytes at 298.15 K against

mole fraction. The PDH contribution is always positive, increases with electrolyte

concentration, and has the same value for all single electrolytes at a given ionic strength

(not shown in figure). In addition to the long range contribution, the short range

contribution is required to calculate the excess enthalpy. As mentioned previously, the

temperature dependence of logarithm of the activity coefficients from the short range

contribution, 𝑙𝑛𝛾𝑖𝑙𝑐 , is determined by 𝜏𝑖𝑗 and 𝐺𝑖𝑗 and their temperature derivatives shown

in Eqs. 3.30 and 3.31. At 298.15 K, the excess enthalpy depends on the Gibbs energy

57
Texas Tech University, Nazir Hossain, August 2018

term, 𝛥𝑔𝑖𝑗 , and the enthalpy term, 𝛥ℎ𝑖𝑗 . The heat capacity term, ∆𝑐𝑝 𝑖𝑗 , has no

contribution to the excess enthalpy at 298.15 K. If the Gibbs energy term is known, then

the excess enthalpy at 298.15 K can be calculated from the enthalpy term or, reversely,

the enthalpy term can be quantified from the excess enthalpy data at 298.15 K. Table 3.1

summarizes the regressed values for ∆ℎ𝑤,𝑐𝑎 and ∆ℎ𝑐𝑎,𝑤 for 60 electrolytes. Virtually all

the values of 𝛥ℎ𝑤,𝑐𝑎 are positive while all the values of 𝛥ℎ𝑐𝑎,𝑤 are negative. The only

exception is aqueous MgSO4 solution which shows the values of 𝛥ℎ𝑤,𝑐𝑎 and ∆ℎ𝑐𝑎,𝑤 near

zero. Note that, for the regression, the standard deviation of excess enthalpy is set to 0.1-

5 J/mol depending on the maximum excess enthalpy. Also given in Table 3.1, the root

mean square errors (RMSE) suggest that the enthalpy term parameters capture the data

very well. For a few cases the parameters cannot be determined properly due to scarcity

of data.

Figure 3.9 presents the calculated excess enthalpy of aqueous NaCl electrolyte

solution at 298.15 K with ∆ℎ𝑤,𝑐𝑎 varied from 0 to 500 and ∆ℎ𝑐𝑎,𝑤 set to a constant

limiting value of 0. Taken from Table 1 for NaCl, the values for the Gibbs energy terms,

∆𝑔𝑤,𝑐𝑎 and ∆𝑔𝑐𝑎,𝑤 are kept constant at 8.845 and -4.534, respectively. The calculated

excess enthalpy shifts downward from the positive PDH excess enthalpy line to negative

excess enthalpy as ∆ℎ𝑤,𝑐𝑎 increases from 0 to 500. Figure 3.10 shows the calculated

excess enthalpy with ∆ℎ𝑤,𝑐𝑎 fixed at 500 and ∆ℎ𝑐𝑎,𝑤 varied from −200 to 0. Again the

calculated excess enthalpy shifts downward from the positive PDH excess enthalpy line

to negative excess enthalpy as ∆ℎ𝑐𝑎,𝑤 increases from -200 to 0.

58
Texas Tech University, Nazir Hossain, August 2018

Figure 3.11 summarizes readily accessible literature heat capacity data for 26

aqueous single electrolytes at 298.15 K [21-26]. Among these electrolytes, both 1-1 and

1-2 electrolytes show similar concentration dependence. The heat capacity of aqueous 1-1

and 1-2 electrolytes decrease at similar rate with increase of electrolyte concentration.

However aqueous 2-1 electrolytes show relatively steeper decrease of the heat capacity

compared to the aqueous 1-1 and 1-2 electrolytes. Figure 3.11 also shows literature heat

capacity data for most of the aqueous 1-1 and 1-2 electrolytes are available at low

concentration region of electrolyte mole fraction ≤ 0.02. The excess heat capacities of

aqueous NaCl, KCl, Na2SO4, K2SO4, CaCl2, and MgCl2 solutions are shown in Figure

3.12. Positive deviations are observed for these selected electrolytes.

The third temperature coefficients of the interaction parameters, ∆𝑐𝑝 𝑖𝑗 , are related

to the excess heat capacity of electrolyte solutions at 298.15 K. To compute the excess

heat capacity, 𝑐𝑝∗𝑒𝑥 , at 298.15 K, all the three temperature coefficients of the interaction

∞,𝑎𝑞
parameters are required. Additionally, the infinite dilution heat capacities of ions, 𝑐𝑝,𝑘 ,

∞,𝑎𝑞
are also required. Reliable literature values for 𝑐𝑝,𝑘 are few [19]. Fortunately, this

quantity can be regressed from experimental calorimetric data [16, 27-30]. Table 3.2

summarizes the thermodynamic constants for water and ions at 298.15 K and

atmospheric pressure which are required to calculate excess enthalpy and excess heat

capacity.

Figures 3.13 and 3.14 show the eNRTL model predictions for the excess heat

capacity of aqueous NaCl solution at 298.15 K. Here the Gibbs energy term and the

enthalpy term are taken from Table 1. Yan and Chen [16] reported from their

59
Texas Tech University, Nazir Hossain, August 2018

comprehensive study that ∆𝑐𝑝 𝑤,𝑐𝑎 and ∆𝑐𝑝 𝑐𝑎,𝑤 for aqueous NaCl electrolyte are -3.251

and 3.11 respectively. In Figure 3.13, ∆𝑐𝑝 𝑤,𝑐𝑎 is kept constant at -3 and ∆𝑐𝑝 𝑐𝑎,𝑤 is varied

from 3 to -3. The excess heat capacity shifts downward to negative as ∆𝑐𝑝 𝑐𝑎,𝑤 decreases

from 3 to -3. Figure 14 shows the excess heat capacity predictions with ∆𝑐𝑝 𝑤,𝑐𝑎 varying

from 3 to -3 and ∆𝑐𝑝 𝑐𝑎,𝑤 fixed at 3. Again, the excess heat capacity shifts downward as

∆𝑐𝑝 𝑤,𝑐𝑎 decreasing from 3 to -3. However, the calculated excess heat capacity remains

positive. The regressed heat capacity terms for selected electrolytes are shown in Table

3.3. It should be noted that, due to the limitation of ASPEN, for the regression purpose

the heat capacity data are used rather than the excess heat capacity. The standard

deviation for the heat capacity is selected as 1%.

In summary, the Gibbs energy, enthalpy, and heat capacity terms of the eNRTL

binary interaction parameters can be quantified by regressing the mean ionic activity

coefficients, excess enthalpy, and excess heat capacity data or their equivalents at 298.15

K. Comprehensive studies on aqueous solution of NaCl [16], Na2SO4 [16], KCl [27],

K2SO4 [28], MgSO4 [29], CaCl2 [30], and BaCl2 [31] have been reported considering

experimental data such as mean ionic activity coefficient, osmotic coefficient, vapor

pressure, heat capacity, enthalpy, etc. over wide ranges of temperature. However,

regression with mean ionic activity coefficients, excess enthalpy, and heat capacity at

298.15 K would yield similar interaction parameters. Figure 3.15 presents the

comparison of the temperature dependence of the interaction parameters between a

comprehensive study [16] and this study for aqueous NaCl and aqueous Na2SO4 single

60
Texas Tech University, Nazir Hossain, August 2018

electrolytes. Note that the parameters may not be well determined for some electrolytes

due to relatively limited amount of excess enthalpy and heat capacity data as well as due

to inherent uncertainties of data. In any case, these three temperature coefficients, when

properly identified, should provide a sound thermodynamic basis to address the

temperature dependence of the interaction parameters.

3.5 Conclusions
The liquid phase nonideality of aqueous single electrolyte solutions is well

described by the eNRTL model and the binary interaction energy parameters with the

temperature dependence expressed in a Gibbs-Helmholtz type equation. The three

temperature coefficients for the binary parameters are related to the excess Gibbs energy,

the excess enthalpy, and the excess heat capacity contribution terms respectively. The

excess Gibbs energy terms and the excess enthalpy terms have been successfully

regressed from data on mean ionic activity coefficient and excess enthalpy of aqueous

single electrolyte solutions at 298.15 K. The excess heat capacity terms for some selected

electrolytes are also quantified from experimental heat capacity data at 298.15 K. The

resulting eNRTL binary parameters including the temperature coefficients show

significant common trends and they should offer a useful guideline for the temperature

dependence of the eNRTL binary interaction parameters in the development of

thermodynamic models for electrolyte systems.

61
Texas Tech University, Nazir Hossain, August 2018

Table 3.1 Regressed Gibbs energy and enthalpy terms of interaction parameters from
mean ionic activity coefficient [17] and excess enthalpy [19] of electrolytes at 298.15 K
with 𝛼 = 0.2

RMSE_ℎ∗𝑒𝑥 b

Electrolytes 𝛥𝑔𝑤,𝑐𝑎 𝛥𝑔𝑐𝑎,𝑤 𝛥ℎ𝑤,𝑐𝑎 𝛥ℎ𝑐𝑎,𝑤 RMSE_ 𝛾±𝑚 a


(J.mol-1)

AgNO3 7.424 −3.287 X X 0.010 X

CsCl 8.533 −4.211 509.4 −52.8 0.006 0.90

CsBr 8.512 −4.189 X X 0.006 X

CsI 8.340 −4.088 X X 0.006 X

CsNO3 8.928 −4.117 X X 0.009 X

KAc 8.581 −4.598 1174.0 −503.1 0.008 11.53

KH2PO4 8.973 −4.138 2344.2 −1041.1 0.003 0.86

KF 8.729 −4.475 439.2 −160.2 0.005 6.26

KCl 8.082 −4.114 450.7 −50.4 0.003 1.63

KBr 8.110 −4.149 455.1 −14.9 0.004 3.86

KI 7.914 −4.115 850.7 −121.9 0.005 11.95

KCNS 7.677 −3.914 949.9 −119.4 0.004 26.45

KNO3 7.807 −3.552 6649.0 −2870.0 0.007 7.46

LiCl 10.030 −5.153 658.3 −279.1 0.038 3.32

LiBr 10.440 −5.345 642.9 −254.0 0.047 7.62

LiI 9.260 −4.990 528.0 −212.9 0.023 0.22

LiClO4 9.581 −5.088 539.5 −206.1 0.022 0.21

LiNO3 9.632 −4.932 810.11 −238.0 0.048 13.47

62
Texas Tech University, Nazir Hossain, August 2018

Table 3.1 Continued

LiOH 9.015 −4.403 379.08 −237.2 0.023 0.91

NaF 8.028 −4.031 X X 0.003 X

NaCl 8.845 −4.534 371.2 −37.3 0.017 1.29

NaBr 8.344 −4.387 476.2 −46.9 0.016 4.06

NaI 8.746 −4.608 616.6 −83.1 0.010 10.27

NaAc 8.279 −4.445 1029.3 −454.8 0.007 152.95

NaClO3 7.639 −3.875 1015.8 −146.6 0.006 3.76

NaClO4 8.106 −4.164 806.29 −47.5 0.010 3.50

NaBrO3 7.954 −3.883 4372.3 −1828.6 0.004 0.08

NaH2PO4 8.271 −3.889 324.2 −57.2 0.004 1.42

NaCNS 7.737 −4.159 1218.8 263.2 0.011 8.21

NaNO3 7.709 −3.800 2669.4 −770.4 0.012 13.42

NH4NO3 7.515 −3.560 920.1 −30.6 0.014 23.67

NH4Cl 7.882 −4.022 128.4 −1.8 0.002 0.45

1-2 electrolytes

Cs2SO4 7.162 −3.619 2410.0 −832.8 0.023 0.38

K2SO4 8.784 −4.279 1095.3 −318.0 0.007 0.34

K2CrO4 7.726 −3.876 1430.0 −394.4 0.018 9.26

Na2S2O3 8.010 −4.000 1273.0 −219.9 0.025 13.87

Na2SO4 7.840 −3.830 1354.7 −338.8 0.019 1.79

63
Texas Tech University, Nazir Hossain, August 2018

Table 3.1 Continued

(NH4)2SO4 8.010 −3.865 694.1 −197.1 0.018 1.27

Li2SO4 7.978 −4.080 1087.7 −457.7 0.019 1.78

Rb2SO4 7.152 −3.572 X X 0.007 X

2-1 electrolytes

BaCl2 7.309 −3.992 1075.3 −369.1 0.018 0.41

BaBr2 8.242 −4.435 X X 0.021 X

CaCl2 10.573 −5.315 722.2 −263.0 0.185 14.16

CaBr2 11.311 −5.624 720.60 −298.6 0.356 0.02

Ca(ClO4)2 11.020 −5.6 701.6 −206.7 0.255 44.32

Ca(NO3)2 8.534 −4.375 1489.0 −344.4 0.052 14.60

CoCl2 9.492 −4.950 759.2 −488.7 0.047 1.15

Cu(NO3)2 9.618 −4.927 1509.2 −399.0 0.099 17.74

MgCl2 10.587 −5.380 802.9 −315.6 0.184 19.94

MgBr2 11.155 −5.629 233.3 −79.0 0.219 0.02

MgI2 11.617 −5.840 785.0 −331.8 0.289 0.02

Mg(NO3)2 9.940 −5.108 892.1 −218.83 0.111 8.10

Mg(ClO4)2 11.145 −5.703 514.4 −167.4 0.186 6.42

MnCl2 8.950 −4.704 1408.3 −659.7 0.037 3.52

NiCl2 10.021 −5.146 382.8 −202.7 0.081 4.13

SrCl2 9.620 −4.949 969.8 −387.8 0.077 0.07

64
Texas Tech University, Nazir Hossain, August 2018

Table 3.1 Continued

SrBr2 8.919 −4.745 549.2 −119.1 0.032 0.04

Sr(NO3)2 7.537 −3.848 4766.4 −1637.4 0.039 10.36

Sr(ClO4)2 10.557 −5.388 484.3 −107.7 0.152 3.34

UO2(NO3)2 8.779 −4.782 4437.9 −1491.9 0.042 1.98

ZnCl2 7.557 −3.968 336.9 −540.0 0.104 36.14

Zn(ClO4)2 11.298 −5.752 507.6 −134.0 0.187 16.80

Zn(NO3)2 9.998 −5.122 2361.3 −687.6 0.131 111.91

2-2 electrolytes

MgSO4 8.987 −4.431 −49.127 24.32 0.054 1.26

BeSO4 9.162 −4.533 196.2 −181.3 0.057 8.43

MnSO4 8.912 −4.368 35.5 −56.3 0.051 6.34

1 𝑍𝑖 −𝑍𝑀𝑖 2
a
RMSE_ 𝛾±𝑚 = √𝑘 ∑𝑘𝑖=1 ( )
𝑍𝑀𝑖

1
b
RMSE_ ℎ∗𝑒𝑥 = √𝑘 ∑𝑘𝑖=1(𝑍𝑖 − 𝑍𝑀𝑖 )2

X- No sufficient data to determine the parameters

65
Texas Tech University, Nazir Hossain, August 2018

Table 3.2 Thermodynamic constants for water and ions at 298.15 K and 0.1 MPa [19]

∆𝑓 𝐻 𝐶𝑝
Component State
(kJ.mol-1) (J.mol-1K-1)

H2O ig −241.82 33.6

K+ ∞, aq −252.38 21.8

Na+ ∞, aq −240.12 46.4

Ca2+ ∞, aq −542.83 −153.8*

Mg2+ ∞, aq −466.85 −90.0

Cl− ∞, aq −131.228 −116.2**

SO42− ∞, aq −909.27 −244.2**

*From [30]
**From [16]

66
Texas Tech University, Nazir Hossain, August 2018

Table 3.3 Regressed heat capacity terms of interaction parameters from heat capacity
data [21-26] at 298.15 K with 𝛼 = 0.2

RMSE Cp a
Electrolytes ∆𝑐𝑝 𝑤,𝑐𝑎 ∆𝑐𝑝 𝑐𝑎,𝑤
(J.mol-1K-1)

NaCl −8.962 5.576 0.07

KCl 0 0.603 0.05

Na2SO4 8.024 −0.992 0.02

K2SO4 −0.019 0.595 0.04

CaCl2 −0.335 2.355 0.37

MgCl2 −5.197 2.250 0.08

1
a
RMSE Cp = √𝑘 ∑𝑘𝑖=1(𝑍𝑖 − 𝑍𝑀𝑖 )2

67
Texas Tech University, Nazir Hossain, August 2018

Figures

Figure 3.1 Temperature dependence of interaction parameters of Pitzer model for NaCl +
H2O and Na2SO4 + H2O binary systems; (a) NaCl + H2O: ( ) 10*β(0), ( ) β(1), (
γ γ
) 100*C ; (b) Na2SO4 + H2O: ( ) 10*β , (
(0)
)β ,(
(1)
) 100*C

68
Texas Tech University, Nazir Hossain, August 2018

Figure 3.2 Temperature dependence of eNRTL binary interaction parameters for NaCl +
H2O and Na2SO4 + H2O binary systems [16]; ( ) H2O - (Na+ Cl-), ( ) (Na+ Cl-) -
H2O, ( ) H2O - (Na+ SO42-), ( ) (Na+ SO42-) - H2O

69
Texas Tech University, Nazir Hossain, August 2018

Figure 3.3 Experimental mole fraction scale mean ionic activity coefficients of 1-1
electrolytes at 298.15 K [17]

70
Texas Tech University, Nazir Hossain, August 2018

Chart Ti
rt Title
1.2
art Title Chart Title Chart Title
1.2
Figure 3.4 Mean ionic activity coefficient of 1-1 electrolytes at 298.15 K calculated by
the PDH term and the lc term with various values for (∆g1 w,ca , ∆g ca,w ): ( ) PDH; (
1 ) (6, −3); ( ) (8, −4); ( ) (8.845, −4.534) (NaCl)); ( ) (10, −6);(
) (12, −6) 0.8
0.8

0.6 0.6
0.4
0.4
0.2

0 0.2
0.2 0 0.4 0.2 0.6 0.4 0.8 0.6 1 0.8 1.2 1 1.2

0
0 0.2 0.4 0.6
0.6
0.6 0.8
0.8 11 1.21.2

71
Texas Tech University, Nazir Hossain, August 2018

Chart Title Chart Title Chart Title


Chart TitleChart Title
1.2 1.2 1.2
Figure 3.5 Symmetric excess Gibbs energy of aqueous 1-1 electrolyte solution at 298.15
K calculated by the PDH term and the lc term with varying values for (∆g w,ca , ∆g ca,w ): (
1 1
1 ) PDH; ( ) (6, −3) ( ) (6.9315, −3.4657); ( ) (8, −4); ( )
(8.845, −4.534) (NaCl)); ( ) (10, −5)
0.8 0.8 0.8

0.6
0.6 0.6
0.4
0.4 0.4
0.2
0.2 0.2
0
0 0.2 0.4 0.6 0.8 1 1.2
0 0
0.4 00.2 0.6 0.20.4 0.80 0.40.6 10.2 0.60.8 0.4 0.81
1.2 0.6 11.2 0.81.2 1 1.2

72
Texas Tech University, Nazir Hossain, August 2018

Figure 3.6 eNRTL model results for symmetric excess Gibbs energy of some aqueous
electrolytes at 298.15 K

73
Texas Tech University, Nazir Hossain, August 2018

Figure 3.7 Experimental excess enthalpy of some 1-1 electrolytes in unsymmetric


reference state at 298.15 K [19]

74
Texas Tech University, Nazir Hossain, August 2018

Figure 3.8 Pitzer-Debye-Hückel prediction of excess enthalpy of aqueous electrolytes at


298.15 K: ( ) 1-1 electrolytes, ( ) 1-2/2-1, ( ) 2-2, ( ) 1-3/3-1

75
Texas Tech University, Nazir Hossain, August 2018

Figure 3.9 eNRTL model prediction of excess enthalpy of aqueous NaCl solution at
298.15 K along with experimental data [19] with fixed values for Gibbs energy terms
(∆g w,ca= 8.845 and ∆g ca,w = −4.534) and varying values for enthalpy terms (∆hw,ca ,
∆hca,w ) of interaction parameters: ( ) PDH; ( ) (100, 0); ( ) (200, 0); (
) (300, 0); ( ) (400, 0) ; ( ) (500, 0)

76
Texas Tech University, Nazir Hossain, August 2018

Figure 3.10 eNRTL model prediction of excess enthalpy of aqueous NaCl solution at
298.15 K along with experimental data [19] with fixed values for Gibbs energy terms
(∆g w,ca= 8.845 and ∆g ca,w = −4.534) and varying values for enthalpy terms (∆hw,ca,
∆hca,w ) of interaction parameters: ( ) PDH; ( ) (500, −200); ( ) (500,
−150); ( ) (500, −100); ( ) (500, −50) ; ( ) (500, 0)

77
Texas Tech University, Nazir Hossain, August 2018

Figure 3.11 Experimental heat capacity data of some 1-1, 1-2, and 2-1 electrolytes at
298.15 K [21-26]

78
Texas Tech University, Nazir Hossain, August 2018

Figure 3.12 Experimental excess heat capacity data at 298.15 K: ( ) NaCl [23], (
) KCl [26], ( ) Na2SO4, [25] ( ) K2SO4, ( ) CaCl2 [23] and ( )
MgCl2 [23]

79
Texas Tech University, Nazir Hossain, August 2018

Figure 3.13 eNRTL model prediction of excess heat capacity of aqueous NaCl solution
at 298.15 K along with experimental data (Δ) [21] (○) [23] with fixed values for Gibbs
energy (∆g w,ca= 8.845 and ∆g ca,w = −4.534) and enthalpy terms (∆hw,ca = 371.2and
∆hca,w = −37.3) and changing values for heat capacity terms (∆𝑐𝑝 𝑤,𝑐𝑎 , ∆𝑐𝑝 𝑐𝑎,𝑤 ) of
interaction parameters: ( ) (−3, 3); ( ) (−3, 2); ( ) (−3, 0) ( ) (−3,
−2); ( ) (−3, −3)

80
Texas Tech University, Nazir Hossain, August 2018

Figure 3.14 eNRTL model prediction of excess heat capacity of aqueous NaCl solution
at 298.15 K along with experimental data (Δ) [21] (○) [23] with fixed values for Gibbs
energy (∆g w,ca= 8.845 and ∆g ca,w = −4.534) and enthalpy terms (∆hw,ca = 371.2and
∆hca,w = −37.3) and changing values for heat capacity terms (∆𝑐𝑝 𝑤,𝑐𝑎 , ∆𝑐𝑝 𝑐𝑎,𝑤 ) of
interaction parameters: ( ) (−3, 3); ( ) (−2, 3); ( ) (0, 3); ( ) (2, 3); (
) (3, 3)

81
Texas Tech University, Nazir Hossain, August 2018

Chart Title
Chart TitleFigure 3.15 Temperature dependence of eNRTL binary interaction parameters for NaCl
Chart Title + H21.2
O and Na2SO4 + H2O binary systems; Comprehensive study [16] ( ) H2O - (Na+
Cl-), ( ) (Na+ Cl-) - H2O, ( ) H2O - (Na+ SO42-), ( ) (Na+ SO42-) - H2O;
+ - + -
This study:
1 ( ) H2O - (Na Cl ), ( ) (Na Cl ) - H2O, ( ) H2O - (Na+ SO42-),
( ) (Na+ SO42-) - H2O
0.8

0.6

0.4

0.2
0.4 0.6 0.8 1 1.2
0
0 0.2 0.4 0.6 0.8 1 1.2
0.6 0.8 1 1.2

82
Texas Tech University, Nazir Hossain, August 2018

References

[1] C.-C. Chen, "Toward Development of Activity Coefficient Models for Process
and Product Design of Complex Chemical Systems," Fluid phase equilibria, vol.
241, pp. 103-112, 2006.
[2] D. L. Shaffer, L. H. Arias Chavez, M. Ben-Sasson, S. Romero-Vargas Castrillón,
N. Y. Yip, and M. Elimelech, "Desalination and Reuse of High-Salinity Shale Gas
Produced Water: Drivers, Technologies, and Future Directions," Environmental
Science & Technology, vol. 47, pp. 9569-9583, 2013.
[3] B. R. Hansen and S. R. Davies, "Review of Potential Technologies for the
Removal of Dissolved Components from Produced Water," Chemical
Engineering Research & Design, vol. 72, pp. 176-188, 1994.
[4] A. D. Khawaji, I. K. Kutubkhanah, and J.-M. Wie, "Advances in seawater
Desalination Technologies," Desalination, vol. 221, pp. 47-69, 2008.
[5] K. S. Pitzer, "Thermodynamics of electrolytes. I. Theoretical Basis and General
Equations," The Journal of Physical Chemistry, vol. 77, pp. 268-277, 1973.
[6] K. Thomsen, "Modeling Electrolyte Solutions with The Extended Universal
Quasichemical (UNIQUAC) Model," Pure and Applied Chemistry, vol. 77, pp.
531-542, 2005.
[7] F. F. Hingerl, T. Wagner, D. A. Kulik, K. Thomsen, and T. Driesner, "A New
Aqueous Activity Model for Geothermal Brines in the System Na-K-Ca-Mg-H-
Cl-SO4-H2O from 25 to 300 °C.," Chemical Geology, vol. 381, pp. 78-93, 2014.
[8] P. Wang, A. Anderko, and R. D. Young, "A Speciation-Based Model for Mixed-
Solvent Electrolyte Systems," Fluid Phase Equilibria, vol. 203, pp. 141-176,
2002.
[9] P. Wang, A. Anderko, R. D. Springer, and R. D. Young, "Modeling Phase
Equilibria and Speciation in Mixed-Solvent Electrolyte Systems: II. Liquid–
Liquid Equilibria and Properties of Associating Electrolyte Solutions," Journal of
Molecular Liquids, vol. 125, pp. 37-44, 2006.
[10] C. C. Chen, H. Britt, J. Boston, and L. Evans, "Local Composition Model for
Excess Gibbs Energy of Electrolyte Systems. Part I: Single Solvent, Single
Completely Dissociated Electrolyte Systems," AIChE Journal, vol. 28, pp. 588-
596, 1982.
[11] C.-C. Chen and L. B. Evans, "A Local Composition Model for the Excess Gibbs
Energy of Aqueous Electrolyte Systems," AIChE Journal, vol. 32, pp. 444-454,
1986.
[12] Y. Song and C.-C. Chen, "Symmetric Electrolyte Nonrandom Two-Liquid
Activity Coefficient Model," Industrial & Engineering Chemistry Research, vol.
48, pp. 7788-7797, 2009.
[13] K. S. Pitzer, Thermodynamics, 3rd ed. New York: McGraw Hill, 1995.
[14] W. Voigt, "Chemistry of Salts in Aqueous Solutions: Applications, Experiments,
and Theory," in Pure and Applied Chemistry vol. 83, ed, 2011, p. 1015.

83
Texas Tech University, Nazir Hossain, August 2018

[15] E. Clarke and D. Glew, "Evaluation of Thermodynamic Functions from


Equilibrium Constants," Transactions of the Faraday Society, vol. 62, pp. 539-
547, 1966.
[16] Y. Yan and C.-C. Chen, "Thermodynamic Representation of the NaCl + Na2SO4
+ H2O System with Electrolyte NRTL Model," Fluid Phase Equilibria, vol. 306,
pp. 149-161, 2011.
[17] R. A. Robinson and R. H. Stokes, Electrolyte Solutions: Courier Dover
Publications, 2002.
[18] Y. Marcus, "Effect of Ions on the Structure of Water: Structure Making and
Breaking," Chemical Reviews, vol. 109, pp. 1346-1370, 2009.
[19] D. D. Wagman, W. H. Evans, V. B. Parker, R. H. Schumm, I. Halow, S. M.
Bailey, et al., "The NBS Tables of Chemical Thermodynamic Properties. Selected
Values for Inorganic and C1 and C2 Organic Substances in SI Units," DTIC
Document 1982.
[20] C.-C. Chen, P. M. Mathias, and H. Orbey, "Use of Hydration and Dissociation
Chemistries with the Electrolyte–NRTL Model," AIChE Journal, vol. 45, pp.
1576-1586, 1999.
[21] J. L. Fortier, P. A. Leduc, and J. E. Desnoyers, "Thermodynamic Properties of
Alkali Halides. II. Enthalpies of Dilution and Heat Capacities in Water at 25 °C,"
Journal of Solution Chemistry, vol. 3, pp. 323-349, 1974.
[22] G. Perron, J. E. Desnoyers, and F. J. Millero, "Apparent Molal Volumes and Heat
Capacities of Alkaline Earth Chlorides in Water at 25 °C," Canadian Journal of
Chemistry, vol. 52, pp. 3738-3741, 1974.
[23] G. Perron, A. Roux, and J. E. Desnoyers, "Heat Capacities and Volumes of NaCl,
MgCl2, CaCl2, and NiCl2 up to 6 Molal in Water," Canadian Journal of
Chemistry, vol. 59, pp. 3049-3054, 1981.
[24] P. P. S. Saluja and J. C. LeBlanc, "Apparent Molar Heat Capacities and Volumes
of Aqueous Solutions of Magnesium Chloride, Calcium Chloride, and Strontium
Chloride at Elevated Temperatures," Journal of Chemical & Engineering Data,
vol. 32, pp. 72-76, 1987.
[25] P. P. S. Saluja, R. J. Lemire, and J. C. LeBlanc, "High-Temperature
Thermodynamics of Aqueous Alkali-Metal salts," The Journal of Chemical
Thermodynamics, vol. 24, pp. 181-203, 1992.
[26] R. T. Pabalan and K. S. Pitzer, "Apparent Molar Heat Capacity and other
Thermodynamic Properties of Aqueous Potassium Chloride Solutions to High
Temperatures and Pressures," Journal of Chemical & Engineering Data, vol. 33,
pp. 354-362, 1988.
[27] S. K. Bhattacharia and C.-C. Chen, "Thermodynamic Modeling of KCl + H2O and
KCl + NaCl + H2O Systems using Electrolyte NRTL Model," Fluid Phase
Equilibria, vol. 387, pp. 169-177, 2015.
[28] S. K. Bhattacharia, N. Hossain, and C.-C. Chen, "Thermodynamic Modeling of
Aqueous Na+–K+–Cl−–SO42− Quaternary System with Electrolyte NRTL Model,"
Fluid Phase Equilibria, vol. 403, pp. 1-9, 2015.

84
Texas Tech University, Nazir Hossain, August 2018

[29] S. K. Bhattacharia, S. Tanveer, N. Hossain, and C.-C. Chen, "Thermodynamic


Modeling of Aqueous Na+–K+–Mg2+–SO42− Quaternary System," Fluid Phase
Equilibria, vol. 404, pp. 141-149, 2015.
[30] S. Tanveer and C.-C. Chen, "Thermodynamic Modeling of Aqueous Ca2+– Na+–
K+–Cl− Quaternary System," Fluid Phase Equilibria, vol. 409, pp. 193-206, 2016.
[31] S. H. Saravi, S. Honarparvar, and C.-C. Chen, "Modeling Aqueous Electrolyte
Systems," Chemical Engineering Progress, vol. 111, pp. 65-75, 2015.

85
Texas Tech University, Nazir Hossain, August 2018

CHAPTER 4
THERMODYNAMIC MODELING OF AQUEOUS LI+–NA+–K+–
MG2+−SO42− QUNIARY SYSTEMS
4.1 Abstract
A comprehensive thermodynamic modeling of aqueous Li+–Na+–K+–Mg2+−SO42−

quinary system has been developed with electrolyte nonrandom two liquid (eNRTL)

model. For this quinary system, the eNRTL model requires four electrolyte-water and six

electrolyte-electrolyte binary interaction parameters. The eNRTL binary interaction

parameters for (Li+ SO42−):H2O, (Li+ SO42−):(Na+ SO42−), (Li+ SO42−):(K+ SO42−), and (Li+

SO42−):(Mg2+ SO42−) pairs are obtained from the correlation of the osmotic coefficient,

excess enthalpy, and salt solubility data. The other binary interaction parameters are

obtained from the literature. The eNRTL model represents various thermodynamic

properties accurately from 273.15–573.15 K and concentration up to saturation.

Keywords: Quinary system; Electrolyte NRTL Model; Lithium Sulfate; Thermodynamic

properties

86
Texas Tech University, Nazir Hossain, August 2018

4.2 Introduction
Hydraulic fracturing has gained popularity in the oil and gas production industries

to increase the production of the crude oil and natural gas [1]. However, this process

yields a very high amount of water, known as “produced” water which contains of oil,

grease, chemicals, suspended organics, naturally occurring radioactive materials, and

total dissolved solids [2]. The major cations in TDS are sodium, potassium, calcium,

magnesium, and lithium, while the major anions are chloride, bicarbonate, and sulfate

[3]. The other chemical processes involve highly concentrated electrolytes are

desalination of the seawater [4-6], nuclear waste treatment [7-9], extraction minerals

from the salt lake etc. [10]. Electrolyte solutions comprise a significant group of natural

and industrial systems [11]. To facilitate thermodynamic property calculations, analysis,

simulation and design of chemical processes involving electrolytes, accurate

thermodynamic modeling of electrolyte solutions is essential [12]. This work aims to

develop a thermodynamic model for aqueous lithium sulfate and integrate with

previously developed model for Na+–K+–Mg2+–SO42− quaternary system [13, 14] as a

continuing effort to develop a comprehensive thermodynamic model for highly saline

water.

Rard et al. [15] developed a Pitzer ion-interaction model for aqueous system

containing lithium sulfate from the isopiestic measurement of the osmotic and activity

coefficients at 298.15 K and 323.15 K and concentration up to saturation.

Filippov et al. [16] measured osmotic coefficients for aqueous lithium sulfate and

cesium sulfate at 298.15 K and calculated the Pitzer model parameters from these

measurements. Additionally, they determined the thermodynamic quantities for the


87
Texas Tech University, Nazir Hossain, August 2018

crystalline compounds. Also, in another study [17], they measured osmotic coefficients of

aqueous lithium sulfate in the presence of alkali metals such as Na, K, and Rb at 298.15

K and calculated the Pitzer model parameters and the thermodynamic quantities for

alkali-metal double sulfates.

Wang and Deng [18] determined the solubility and the density in the aqueous

ternary system (Li2SO4 + MgSO4 + H2O) at 308.15 K by isothermal evaporation. Based

on the extended HW (Harvie and Weare) model [19] of aqueous electrolyte solutions, a

variant of the Pitzer model, they calculated the solubility of this ternary system at 308.15

K.

Recently, Zhou et al.[20] , determined the solubility of Li2SO4 + MgSO4 ternary

and the Li2SO4 + K2SO4 + MgSO4 quaternary systems in water at 273.15 K. They used

Pitzer-Simonson-Clegg (PSC) [21, 22] model for representing the properties of the binary

and ternary subsystems of this quaternary system. Seven binary interaction parameters

were required for each aqueous single electrolyte binary and two additional ternary

interaction parameters were fitted for each aqueous two electrolytes ternary at 373.15 K.

The PSC model has been applied to predict the solubility isotherms of the quaternary

system and compared satisfactorily with the experimental results. However, this model

only covers only 273.15 K temperature thus the extension of this model to other

temperatures would be challenging because the Pitzer model and its variants do not

provide any information about the temperature dependency of the model parameters [9,

11] and for each Pitzer parameter, up to eight temperature coefficients would be required

for calculations at temperatures from 273 to about 473 K [23, 24].

88
Texas Tech University, Nazir Hossain, August 2018

Although computer calculations enable the estimation of the large number of

parameters required by Pitzer’s model, the limitation of this model is mainly due to the

scarcity of reliable experimental data for parameter estimation [11]. Therefore, models

that require less number of parameters are more favorable for thermodynamic modeling

of mixed salt solutions [11] . One of the most popular approaches for development of

models with fewer number of parameters compared to the Pitzer’s model, is combining a

nonelectrolyte model with an extended Debye–Hückel term. In electrolyte NRTL

(eNRTL) model, the nonelectrolyte model is NRTL [11, 24] .

Since in the eNRTL model, the number of required binary interaction parameters are very

few, it is a practical thermodynamic framework for aqueous electrolyte systems [25] and

is capable of representing the liquid-phase nonideality at a wide concentration interval

from zero to saturation [26]. Furthermore, the model binary interaction parameters are

explicitly expressed in terms of temperature coefficients corresponding to heat capacity,

enthalpy and excess Gibbs free energy contributions of the Gibbs-Helmholtz expression

[25] . Properly identified from thermodynamic data, the temperature coefficients enable

predictions of thermodynamic properties at different temperatures.

This work extends our prior thermodynamic modeling work using the eNRTL

model for aqueous quaternary system Na+–K+–Mg2+–SO42− [14] to the aqueous Li+−Na+–

K+–Mg2+−SO42− quinary system. The binary interaction parameters for Li2SO4 + H2O

binary and ternary systems, which involve Li2SO4, are correlated with osmotic

coefficient, excess enthalpy, and salt solubility data. The other necessary salt-water and

salt-salt binary parameters are adopted from our prior works [6, 13, 14]. The other

89
Texas Tech University, Nazir Hossain, August 2018

thermodynamic constants for ions, water, and precipitating salts are listed in Table 4.3

and Table 4.4 with their sources [6, 13, 14, 27]. All the regressions and the property

calculations are performed in ASPEN V8.4. The resulting eNRTL model correlates

experimental data and able to predict thermodynamic properties

4.3 Thermodynamic Frameworks


Since all the electrolytes are considered as non-volatile only water participates in

the vapor-liquid equilibrium. Formulation of vapor-liquid phase equilibria starts with

equality of the fugacity in the two phases:

𝑃𝜑𝑖 = 𝑥𝑖 𝛾𝑖 𝑓𝑖𝑜 (4.1)

where, 𝑃 is the system pressure, 𝜑𝑖 is the fugacity coefficient of component i at vapor

phase, is calculated from Redlich-Kwang equation of state [], 𝑥𝑖 is the liquid phase mole

fraction, 𝛾𝑖 , the liquid activity coefficient of component i calculated from the eNRTL

model, and 𝑓𝑖𝑜 the liquid phase reference fugacity.

The rationale and detail of eNRTL model are discussed elsewhere [6, 24]. In

short, the excess Gibbs free energy, 𝐺 𝑒𝑥 , of the aqueous electrolytes can be calculated

from the long range ion-ion and the short range ion-ion, ion-water, and water-water

interactions.

𝐺 𝑒𝑥 = 𝐺 𝑒𝑥,𝑃𝐷𝐻 + 𝐺 𝑒𝑥,𝑙𝑐 (4.2)

where 𝐺 𝑒𝑥,𝑃𝐷𝐻 accounts the long range ion-ion interactions calculated form Pitzer–

Debye–Hückel formula and 𝐺 𝑒𝑥,𝑙𝑐 accounts the short range ion-ion, ion-water, and water-

water interactions calculated form the non-random two liquid (NRTL) theory.

90
Texas Tech University, Nazir Hossain, August 2018

The activity coefficient of species i, 𝛾𝑖 can be calculated from the excess Gibbs

free energy as follows:

1 𝜕𝐺 𝑒𝑥
𝑙𝑛𝛾𝑖 = 𝑅𝑇 ( 𝜕𝑛 ) (4.3)
𝑖 𝑇,𝑃,𝑛𝑗≠𝑖

The Pitzer–Debye–Hückel formula has no adjustable parameters while the NRTL

part requires binary interaction parameters. The interaction parameters of eNRL model

(𝜏𝑖𝑗 ≠ 𝜏𝑗𝑖 ) are expressed as Gibbs-Helmholtz type expression and the temperature

coefficient of the the interaction parameters are adjusted from experimental data [24, 25].

For solid-liquid equilibrium of individual precipitating crystalline salts, solubility

product constant Ksp is used for equilibrium calculations.

𝐾𝑠𝑝 = ∏𝑖(𝑥𝑖𝑠𝑎𝑡 𝛾𝑖 )𝜈𝑖 (4.4)


∆𝐺𝑘𝑜
𝑙𝑛 𝐾𝑠𝑝 (𝑇) = − (4.5)
𝑅𝑇
𝑜
∆𝐺𝑘𝑜 ∆𝐺𝑘𝑜 (𝑇𝑟𝑒𝑓 ) ∆𝐻𝑘𝑜 (𝑇𝑟𝑒𝑓 ) 1 1 ∆𝐶𝑝,𝑘 (𝑇𝑟𝑒𝑓 ) 𝑇𝑟𝑒𝑓 𝑇
=− + [𝑇 − 𝑇] + [ − 1 + 𝑙𝑛 (𝑇 )] (4.6)
𝑅𝑇 𝑅𝑇𝑟𝑒𝑓 𝑅 𝑟𝑒𝑓 𝑅 𝑇 𝑟𝑒𝑓

where 𝑥𝑖𝑠𝑎𝑡 is the mole fraction of component i at saturation, 𝜈𝑖 the reaction

stoichiometric coefficient, ∆𝐺𝑘𝑜 , ∆𝐻𝑘𝑜 , and ∆𝐶𝑝,𝑘


𝑜
the free energy, enthalpy, and heat

capacity change of formation of crystal k, respectively. The reference temperature 𝑇𝑟𝑒𝑓 is

taken as 298.15 K.

4.4 Li2SO4 + H2O Binary


For aqueous lithium sulfate binary, L2SO4 dissociates completely as follows:

𝐿𝑖2 𝑆𝑂4 → 2𝐿𝑖 + + 𝑆𝑂42− (4.7)

At saturation, 𝐿𝑖 + and 𝑆𝑂42− forms a stable salt 𝐿𝑖2 𝑆𝑂4 . 𝐻2 𝑂(𝑐𝑟) by following:

2𝐿𝑖 + (𝑎𝑞) + 𝑆𝑂42− (𝑎𝑞) + 𝐻2 𝑂(𝑎𝑞) ↔ 𝐿𝑖2 𝑆𝑂4 . 𝐻2 𝑂(𝑐𝑟) (4.8)

91
Texas Tech University, Nazir Hossain, August 2018

Limited experimental thermodynamic data for aqueous Li2SO4 binary system are

available and they are primarily at 298.15 K. Recently, Rard et al. [15] compiled and

critically assessed the published osmotic coefficient and mean ionic activity coefficient

data at 298.15 K and 323.15 K. Other thermodynamic data include vapor pressure, excess

enthalpy, heat capacity, and salt solubility are summarized in Table 4.1, along with the

osmotic coefficient and mean ionic activity coefficient data with temperature and

concentration ranges.

In regression, the osmotic coefficient reported by Rard ert al. [15] , excess

enthalpy [27], and salt solubility data [28] are used. For osmotic coefficient data, the

standard deviation in osmotic coefficient was set to 5% and temperature, pressure, and

mole fraction of L2SO4 were set to error free. We excluded four data points (two for each

temperature) were excluded for the regression since they seemed to be outliers. For

excess enthalpy, the standard deviation for excess enthalpy was 0.01 J/mol while

temperature, pressure, and mole fraction of L2SO4 were set to error free. The standard

deviation in temperature and mole fraction of L2SO4 were set to 0.1 K and 5%,

respectively, for the solubility data. All other data listed in Table 4.1 are presented for the

validation purposes.

The Mean relative deviation (MRD) is a measure of agreement between results of

model and experimental data that is calculated via Eq. 4.9

1 𝐸𝑠𝑡−𝐸𝑥𝑝
𝑀𝑅𝐷 (%) = 𝑘 ∑𝑘𝑖=1 | | × 100 (4.9)
𝐸𝑥𝑝

where, Est and Exp are estimated and experimental values and k is the number of

data points. The MRD values for the data used in this model are shown in Table 4.1.

92
Texas Tech University, Nazir Hossain, August 2018

Table 4.2 compiled the regressed eNRTL binary interaction parameters for

Li2SO4 + H2O binary obtained in this work. The thermodynamic constants for the

precipitating slats are shown in the Table 4.3.

Figure 4.1 represents the comparison of the experimental osmotic coefficient

with the eNRTL model prediction at various temperatures. The model captures the

experimental osmotic coefficient data at 298.15 K and 323.15 K [15] satisfactorily with

MRD of 1.39 %. The model gives same MRD for another data source [29] which were

not included in the regression . Figure 4.1 also represents the prediction of the osmotic

coefficient at higher temperatures with a reasonable trend.

The mean ionic activity coefficient of Li2SO4 + H2O binary is predicted with the eNRTL

model. Figure 4.2 shows the comparison of the experimental mean ionic activity

coefficient data [15, 29] with the model predictions. The model predicts the mean ionic

activity coefficient very well, although the data were not used in the model

parameterization. The MRD for the mean ionic activity coefficients for Rard et al. [15]

and Robinson et al. [29] are 1.68 % and 3.37 %, respectively.

Figure 4.3 compares the experimental excess enthalpy data at 298.15 K with the

model results. An excellent agreement with the experimental data and the model results is

achieved with a MRD of 4.01 %.

The predicted vapor pressure are shown with the experimental measurements [30]

at various temperatures in Figure 4.4. The model predicts the vapor pressure of the

Li2SO4 + H2O binary with an excellent accuracy with MRD of 1.44 %.

93
Texas Tech University, Nazir Hossain, August 2018

Figure 4.5 shows the comparison of the experimental heat capacity data [30] with

the eNRTL model predictions. The model capture the trend of the heat capacity at various

temperatures. The MRD of the heat capacity is found to be 1.39 %. It should be noted

that the third temperature coefficient of the interaction parameters of the eNRTL

model, 𝐸𝑖𝑗 , is directly correlated with excess heat capacity. We set the third coefficient to

zero for the simplicity. If we could use the heat capacity data for the model

parameterization, the model would give better prediction.

Figure 4.6 represents the comparison of the solubility of Li2SO4.H2O(s)

temperature 273.15 -500.15 K. The Figure shows an excellent agreement with

experimental data. Shown in Table 4.1, the MRD of the solubility is 2.79 %.

4.5 Ternary Systems


The interaction parameters for the Li2SO4 + H2O binary can be used for the

ternary systems. In this study, we have studied three ternary system which involve

Li2SO4: Li2SO4 + Na2SO4 + H2O, Li2SO4 + K2SO4 + H2O, and Li2SO4 + MgSO4 + H2O.

Table 4.5 shows the compiled experimental solubility data for the systems that include

Li2SO4, K2SO4, Na2SO4, and MgSO4.

4.5.1 Li2SO4 + Na2SO4 + H2O Ternary


In Li2SO4 + Na2SO4 + H2O ternary, Li2SO4 dissociates completely to Li+ and

SO42- ions and Na2SO4 dissociates completely to Na+ and SO42- ions. The ions form

Li2SO4.H2O(s), Li2SO4.3Na2SO4.12H2O(s), and Na2SO4.10H2O(s) at 300.15 K while the

ions precipitate to Li2SO4.H2O(s), Li2SO4.Na2SO4(s), Li2SO4.3Na2SO4.12H2O(s), and

Na2SO4(s) are formed at 318.15 K [28]. The binary interaction parameters for the eNRTL

94
Texas Tech University, Nazir Hossain, August 2018

model of the (Li+ SO42−):(H2O) pair are taken from this study while the (Na+

SO42−):(H2O) pair are taken form literature [6]. The salt-salt binary interactions

parameters for (Li+ SO42−):(Na+ SO42−) pair are regressed from the salt solubility data

[28]. We also regressed the thermodynamic parameters for double salts, Li2SO4.Na2SO4(s)

and Li2SO4.3Na2SO4.12H2O(s) from the salt solubility data. In regression, the standard

deviation in the mole fraction of the precipitating salt are set to 5% while the other salt

mole fraction and temperature are kept error free. If there is a double salt, the standard

deviation of both salts are set to 5%. The identified binary interaction parameters and

thermodynamic parameters for the double salts are listed in Table 4.2 and Table 4.3,

respectively.

Figure 4.7a and Figure 4.7b depict the comparison of the experimental solubility

data of Li2SO4 + Na2SO4 + H2O ternary with the model predictions at 300.15 K and

318.15 K, respectively. The model represents the experimental data reasonably well

including the invariant points with a MRD of 1.4%.

4.5.2 Li2SO4 + K2SO4 + H2O Ternary


Li2SO4 and K2SO4 dissociate completely in water to form the ions Li+, K+, and

SO42-. The stable salts that have been identified for this system are K2SO4(s), Li2SO4.

K2SO4(s), and Li2SO4.H2O(s) at 298.15 K [31]. The binary interaction parameters for the

eNRTL model of the (Li+ SO42−):H2O pair are taken from this study and the (K+

SO42−):(H2O) pair are taken form literature [13]. The only additional parameters that need

to be identified for this system are the eNRTL binary parameters of (Li+ SO42−):(K+

SO42−) pair that are regressed from the experimental solubility data [31] . For Li2SO4.

95
Texas Tech University, Nazir Hossain, August 2018

K2SO4(s), the thermodynamic constant, ∆𝑓 𝐺𝑘𝜊 , is regressed from the experimental

solubility data as well while the other thermodynamic constant, ∆𝑓 𝐻𝑘𝜊 , is taken from the

literature [27] . In the regression the standard deviations are set similar as Li2SO4 +

Na2SO4 + H2O i.e. 5% for the precipitating salts, while error free for non- precipitating

salt and temperature.

Figure 4.8 shows the comparison of the experimental solubility data with model

results at 298.15 K. The model results are in a good agreement with the experimental

data. The MRD for this data is 6%. Figure 4.8 also shows the predictions of the

solubilities at 323.15 K and 348.15 K. The model predictions show the correct trend.

4.5.3 Li2SO4 + MgSO4 + H2O Ternary


In aqueous phase, Li2SO4 and MgSO4 dissociate completely and form Li+, Mg2+,

and SO42- ions. Upon saturation, the ions form Li2SO4·H2O(s) and MgSO4·7H2O(s). The

binary interaction parameters for (Li+ SO42−):H2O pair (Mg2+ SO42−):H2O are already

identified in this study and by Tanveer & Chen [32]. The binary salt-salt interaction

parameters of and (Li+ SO42−):(Mg2+ SO42−) pair are identified form the experimental

solubility data [18, 20]. It should be noted that there is no double salt is formed in this

ternary. In the regression, the standard deviation of the precipitating salt is kept 5%, and

non-precipitating salt and temperature were set error free.

Figure 4.9a and Figure 4.9b represent the solubility diagram of the Li2SO4 +

MgSO4 + H2O ternary at 273.15 K and 308.15 K. The model results are in a very good

agreement with experimental data. The MRD of data reported by Zhou et al. [20] and

Wang and Deng [18] are 3.6% and 3.4%, respectively.

96
Texas Tech University, Nazir Hossain, August 2018

4.6 Quaternary System


The identified binary interaction parameters of the eNRTL model from the binary

and ternary systems can be extended to predict the solubility diagram for the quaternary

systems. We predicted the solubility diagrams Li2SO4 + K2SO4 + MgSO4 + H2O at

273.15 K and compared with the experimental data.

4.6.1 Li2SO4 + K2SO4 + MgSO4 + H2O Quaternary


Zhou [20] measured the solubility data of Li2SO4 + K2SO4 + MgSO4 + H2O

quaternary system at 273.15 K. This quaternary forms five salts, K2SO4(s), Li2SO4·K2SO4

(s), Li2SO4·H2O(s), K2SO4·MgSO4·6H2O(s), and MgSO4·7H2O(s). Figure 4.10 represents

the comparison of the experimental data with the eNRTL model prediction. The model

results show an excellent agreement with the experimental measurements.

4.7 Conclusions
The nonideal behavior of electrolyte solutions, make it challenging to develop an

accurate thermodynamic model for them. Here, for the aqueous quinary system Li+–Na+–

K+–Mg2+–SO42–, a rigorous and comprehensive thermodynamic model has been

developed using eNRTL model that requires only binary interaction parameters

determined from lower order subsystems. From this model, an accurate and

thermodynamically consistent representation of all thermophysical properties such as

mean ionic activity coefficient, osmotic coefficient, vapor pressure, solubility, excess

enthalpy, and heat capacity is obtained that covers concentrations from infinite dilution to

saturation and in the temperature range of 298.15 K to 573.15 K. The model should be

very useful for thermodynamic property calculations and process simulation and

development of novel treatment processes for the aqueous Li+–Na+–K+–Mg2+–SO42–


97
Texas Tech University, Nazir Hossain, August 2018

quinary solution and could be extended to aqueous multicomponent electrolyte systems

containing chlorides and carbonates.

98
Texas Tech University, Nazir Hossain, August 2018

Tables

Table 4.1 Experimental thermodynamic property data for Li2SO4 + H2O binary.

No of
Temperature m (mol.kg- MRD
Data type 1 Data Reference
(K) ) (%)
points
Zaytsev & Aseyev
Vapor pressureb 298.15-573.15 0.47- 2.56 47 1.44
[30]
Osmotic coefficientb 298.15 0.10-3.0 14 Robinson et al. [29] 1.39
Osmotic coefficienta 298.15-323.15 0.11-2.82 124 (120) Rard et al. [15] 1.39
Mean ionic activity
298.15 0.10-3.0 14 Robinson et al.[29] 3.37
coefficientb
Mean ionic activity
298.15-323.15 0.01-3.13 50 Rard et al. [15] 1.68
coefficientb
Excess enthalpya 298.15 0- 3.08 25 Wagman et al. [27] 4.01
Zaytsev & Aseyev
Heat capacityb 298.15-372.15 0-0.03 57 1.36
[30]
Solubilitya 273.15-505.95 2.66- 3.28 36 (34) Seidell [28] 2.79

a
Used in regression.
b
Used for only comparison

99
Texas Tech University, Nazir Hossain, August 2018

Table 4.2 eNRTL binary interaction parameter, τij, for aqueous Li+-Na+-K+-Mg2+-SO42- quinary system.

Component i Component j 𝐶𝑖𝑗 𝐷𝑖𝑗 𝐸𝑖𝑗 a


τij at 298.15 K Reference

H2O (Li+ SO42−) 5.035 ± 0.168 947.2 ± 46.9 0 8.212 ± 0.233 This work
(Li+ SO42−) H2O −2.811 ± 0.069 −401.1 ± 19.4 0 −4.157 ± 0.095 This work
+ 2−
H2O (Na SO4 ) 2.325 1695.5 12.020 8.012 [6]
(Na+ SO42−) H2O −2.208 −505.5 −1.837 −3.903 [6]
H2O (K+ SO42−) 4.045 1354.6 −15.297 8.590 [13]
(K+
SO42−) H2O −2.742 −419.8 8.202 −4.150 [13]
2−
H2O 2+
(Mg SO4 ) 8.244 215.4 −17.325 8.967 [32]
(Mg2+ SO42−) H2O −4.130 −92.2 7.383 −4.438 [32]
(Li+ SO42−) (Na+ SO42−) −3.670 ± 1.240 1190.4 ± 376.6 0 0.323 ± 1.780 This work
(Na+ SO42−) (Li+ SO42−) 2.211 ± 0.317 −760.1 ± 87.975 0 −0.338 ± 0.439 This work
(Li+ SO42−) (K+ SO42−) 0.170 ± 1.269 0 0 0.170 ± 1.269 This work
(K+
SO42−) +
(Li SO4 )2−
−0.782 ± 0.213 0 0 −0.782 ± 0.213 This work
2− 2−
+
(Li SO4 ) 2+
(Mg SO4 ) 3.083 ± 0.947 −892.5 ± 285.7 0 0.090 ± 1.369 This work
(Mg2+ SO42−) (Li+ SO42−) −0.593 ± 0.455 214.9 ± 126.8 0 0.127 ± 0.632 This work
(Na+ SO42−) (K+ SO42−) −0.133 72.6 0 0.111 [13]
(K+ SO42−) (Na+ SO42−) 0.082 −112.6 0 −0.296 [13]
2− 2−
+
(Na SO4 ) 2+
(Mg SO4 ) −0.318 0 0 −0.318 [32]
2− 2−
2+
(Mg SO4 ) +
(Na SO4 ) −1.515 624.7 0 0.580 [32]
(K+ SO42−) (Mg2+ SO42−) −0.949 47.4 0 −0.790 [32]

100
Texas Tech University, Nazir Hossain, August 2018

Table 4.2 Continued

(Mg2+ SO42−) (K+ SO42−) 0.197 0 0 0.197 [32]

𝐷𝑖𝑗 𝑇𝑟𝑒𝑓 −𝑇 𝑇
a
𝜏𝑖𝑗 = 𝐶𝑖𝑗 + + 𝐸𝑖𝑗 ( + 𝑙𝑛 )
𝑇 𝑇 𝑇𝑟𝑒𝑓

101
Texas Tech University, Nazir Hossain, August 2018

Table 4.3 Thermodynamic constants for salts at 298.15 K and 0.1 MPa.

𝜊
∆𝑓 𝐺𝑘𝜊 (kJ.mol-1) ∆𝑓 𝐻𝑘𝜊 (kJ.mol-1) ∆𝑓 𝐶𝑝,𝑘 (J.mol-1K-1)
Salt

Expa Est Expa Est Expa Est


K2SO4 −1321.37 −1321.61b −1437.79 −1439.26b 131.46 ―
Li2SO4.H2O −1565.50 −1566.55c −1735.50 −1740.45c 151.08 ―
Na2SO4 −1270.16 −1269.98 d
−1387.08 −1387.79 d
128.20
Na2SO4∙10H2O −3646.85 −3647.05 d
-4327.26 −4330.03 d
601.43d
MgSO4∙7H2O −2871.50 −2870.35e −3388.71 −3385.92e 385.76* ―
3Na2SO4∙Li2SO4.12H2O ― −7986.15c ― −6630.63c ― ―
K2SO4∙MgSO4.6H2O ― −3957.45 e
― −4545.17 e
479.57 **

K2SO4∙Li2SO4 ― −2656.70c −2883.99 ― ― ―
Li2SO4∙Na2SO4 ― −2614.64c ― −3021.42c ― ―

a
Wagman [27]
b
Bhattacharia et al. [13]
c
This work
d
[6]
e
Tanveer & Chen [32]
*
values estimated by adding corresponding values of MgSO4·6H2O(cr) and H2O(cr)
**
values estimated by adding corresponding values of MgSO4·6H2O(cr), K2SO4(cr), and H2O(cr)

102
Texas Tech University, Nazir Hossain, August 2018

Table 4.4 Thermodynamic constants for water and ions at 298.15 K and 0.1 MPaa.

Component i State ∆𝑓 𝐺 (kJ.mol-1) ∆𝑓 𝐻 (kJ.mol-1) Cp (J mol -1K-1)


H2O ig -228.57 -241.82 33.6
Li+ ∞, aq -293.31 -278.49 68.6
Na+ ∞, aq -261.91 -240.12 46.4
Mg2+ ∞, aq -254.80 -466.85 -18.2b, c
K+ ∞, aq -283.27 -252.38 21.8
SO42− ∞, aq -744.53 -909.27 -244.2d

a
Wagman [27]
b
Tanveer & Chen [32]
∞,𝑎𝑞 𝐶 𝐶 𝐶6
c
𝐶𝑝,𝑘 (𝐽𝑘𝑚𝑜𝑙 −1 𝐾 −1 ) = 𝐶1 + 𝐶2 𝑇(𝐾) + 𝐶3 ((𝑇(𝐾))2 + 𝑇(𝐾)
4 5
+ (𝑇(𝐾)) 2 + ; For Mg2+ ion,
√𝑇(𝐾)
𝐶1 = 191886.16, 𝐶2 = −1843.99, 𝐶3 = 4.21, 𝐶4 = 0, 𝐶5 = 0, 𝐶6 = 0; for SO42− ion 𝐶1 =
−2491007.40, 𝐶2 = 13694.712, 𝐶3 = −20.657, 𝐶4 = 0, 𝐶5 = 0, 𝐶6 = 0;
d
Yan and Chen [6]

103
Texas Tech University, Nazir Hossain, August 2018

Table 4.5 Experimental solubility data for the Li2SO4 in aqueous solution in presence of
other salts. The bullets show what salts are present in the solution.

Composition
Temperature Data
Reference
Li2SO4 Na2SO4 K2SO4 MgSO4 (K) points

● ● 300.15 17 [28]a
● ● 318.15 27 [28] a
● ● 298.15 17 [31]a
● ● 273.15 14 [20]a
● ● 308.15 18 [18]a
● ● ● 273.15 21 [20]b
● ● ● 273.15 21 [10]b

a
Used in regression.
b
Used for comparison.

104
Texas Tech University, Nazir Hossain, August 2018

Figures

Figure 4.1 Osmotic coefficient of aqueous Li2SO4 solution: The solid lines show the
model results compared to the experimental data: at 298.15 K [15] shown in symbols: (□)
298.15 K, (○) 323.15 K, ( ) 298.15 K , ( ) 323.15 K, ( ) 348.15 K, ( )
373.15 K, ( ) 423.15 K.

105
Texas Tech University, Nazir Hossain, August 2018

Figure 4.2 Molality scale mean ionic activity coefficient of Li2SO4 in aqueous Li2SO4
solution: The lines show the model results ( ) 298.15 K, ( ) 323.15 K, ( )
348.15 K, ( ) 373.15 K, ( ) 423.15 K. Experimental data [15] are available at
298.15 K, and 323.15 K which are shown in symbols (□) and (○), respectively.

106
Texas Tech University, Nazir Hossain, August 2018

Figure 4.3 Excess enthalpy of aqueous Li2SO solution: The solid lines show the model
results compared to the experimental data at 298.15 K [27] shown in symbols (○).

107
Texas Tech University, Nazir Hossain, August 2018

Figure 4.4 Vapor Pressure of aqueous Li2SO4 solution: The solid lines show the model
results compared to the experimental data shown in symbols [30] (×) 298.15 K, (◊)
323.15 K, (♦) 338.15 K, (○) 343.15 K, (▲) 363.15 K, (Δ) 373.15 K, (■) 423.15 K, (□)
573.15 K.

108
Texas Tech University, Nazir Hossain, August 2018

Figure 4.5 Heat Capacity of aqueous Li2SO4 solution: The solid lines show the model
results compared to the experimental data shown in symbols [30] (○) 298.15 K, (Δ)
323.15 K, (◊) 348.15 K (×) 363.15 K, (□) 372.15 K.

109
Texas Tech University, Nazir Hossain, August 2018

Figure 4.6 Solubility of Li2SO4 in water: The solid line show the model results compared
to the experimental data shown in symbols (○) [28].

110
Texas Tech University, Nazir Hossain, August 2018

Figure 4.7 Solubility of aqueous Li+–Na+–SO42− ternary system: The solid lines show the
model results compared to the experimental data [28] shown in symbols:
a) Solubility at 300.15 K: (Δ) Li2SO4.H2O, (○) Li2SO4.3Na2SO4.12H2O, (×)
Na2SO4.10H2O
b) Solubility at 318.75 K: (Δ) Li2SO4.H2O, (+) Li2SO4.Na2SO4, (○)
Li2SO4.3Na2SO4.12H2O, (×) Na2SO4

111
Texas Tech University, Nazir Hossain, August 2018

Figure 4.8 Solubility of aqueous Li+–K+–SO42− ternary system: The solid lines show the
model results compared to the experimental data [31] shown in symbols: solubility at
298.15 K: (○) Li2SO4∙H2O, (◊) Li2SO4∙K2SO4, (×) K2SO4. ( ) and ( ) are model
2−
predictions of solubility of aqueous Li –K –SO4 ternary system at 323.15 K and
+ +

348.15 K, respectively.

112
Texas Tech University, Nazir Hossain, August 2018

Figure 4.9 Solubility of aqueous Li+–Mg2+–SO42− ternary system: The solid lines show the
model results compared to the experimental data shown in symbols:
a) Solubility at 273.15 K (○) MgSO4·7H2O, (Δ) Li2SO4·H2O [20]
b) Solubility at 308.15 K (○) MgSO4·7H2O, (Δ) Li2SO4·H2O [18]

113
Texas Tech University, Nazir Hossain, August 2018

Figure 4.10 Jãnecke phase diagram of aqueous Li2SO4 + K2SO4 + MgSO4 + H2O
quaternary system at 273.15 K: The model results and experimental data [20] are shown
with solid lines and symbols (○), respectively.

114
Texas Tech University, Nazir Hossain, August 2018

References
[1] M. Fellet, M. Marder, and T. Patzek, "Science of Hydraulic Fracturing Contains
Materials Questions," MRS Bulletin, vol. 39, p. 484, 2014.
[2] A. Fakhru’l-Razi, A. Pendashteh, L. C. Abdullah, D. R. A. Biak, S. S. Madaeni,
and Z. Z. Abidin, "Review of Technologies for Oil and Gas Produced Water
Treatment," Journal of Hazardous Materials, vol. 170, pp. 530-551, 2009.
[3] E. T. Igunnu and G. Z. Chen, "Produced Water Treatment Technologies,"
International Journal of Low-Carbon Technologies, vol. 9, pp. 157-177, 2014.
[4] K. V. Reddy and N. Ghaffour, "Overview of the Cost of Desalinated Water and
Costing Methodologies," Desalination, vol. 205, pp. 340-353, 2007.
[5] N. Ghaffour, T. M. Missimer, and G. L. Amy, "Technical Review and Evaluation
of the Economics of Water Desalination: Current and future challenges for better
water supply sustainability," Desalination, vol. 309, pp. 197-207, 2013.
[6] Y. Yan and C.-C. Chen, "Thermodynamic Representation of the NaCl + Na2SO4
+ H2O System with Electrolyte NRTL model," Fluid Phase Equilibria, vol. 306,
pp. 149-161, Jul 25 2011.
[7] M. Benedict, H. Levi, and T. Pigford, "Nuclear Chemical Engineering," 1982.
[8] D. D. Sood and S. K. Patil, "Chemistry of Nuclear Fuel Reprocessing: Current
status," Journal of Radioanalytical and Nuclear Chemistry, vol. 203, pp. 547-
573,1996.
[9] M. Wang, M. B. Gorensek, and C.-C. Chen, "Thermodynamic Representation of
Aqueous Sodium Nitrate and Nitric Acid Solution with Electrolyte NRTL
Model," Fluid Phase Equilibria, vol. 407, pp. 105-116, 2016.
[10] Y. Zeng, X. Lin, and X. Yu, "Study on the Solubility of the Aqueous Quaternary
System Li2SO4 + Na2SO4 + K2SO4 + H2O at 273.15 K," Journal of Chemical &
Engineering Data, vol. 57, pp. 3672-3676, 2012.
[11] W. Voigt, "Chemistry of Salts in Aqueous Solutions: Applications, Experiments,
and Theory," Pure and Applied Chemistry, vol. 83, pp. 1015-1030, 2011.
[12] S. H. Saravi, S. Honarparvar, and C.-C. Chen, "Modeling Aqueous Electrolyte
Systems," Chemical Engineering Progress, vol. 111, pp. 65-75, 2015.
[13] S. K. Bhattacharia, N. Hossain, and C.-C. Chen, "Thermodynamic Modeling of
Aqueous Na+–K+–Cl−–SO42− Quaternary System with Electrolyte NRTL Model,"
Fluid Phase Equilibria, vol. 403, pp. 1-9, 2015.
[14] S. K. Bhattacharia, S. Tanveer, N. Hossain, and C.-C. Chen, "Thermodynamic
Modeling of Aqueous Na+–K+–Mg2+–SO42− Quaternary System," Fluid Phase
Equilibria, vol. 404, pp. 141-149, 2015.
[15] J. A. Rard, S. L. Clegg, and D. A. Palmer, "Isopiestic Determination of the
Osmotic and Activity Coefficients of Li2SO4 (aq) at T=298.15 and 323.15 K, and
Representation with an Extended Ion-Interaction (Pitzer) Model," Journal of
Solution Chemistry, vol. 36, pp. 1347-1371, December 01 2007.
[16] V. K. Filippov, A. M. Kalinkin, and S. K. Vasin, "Thermodynamics of Phase
Equilibria of Aqueous (Lithium Sulfate + Cesium Sulfate), (Sodium Sulfate +
Cesium Sulfate), and (Potassium Sulfate + Cesium Sulfate) at 298.15 K using

115
Texas Tech University, Nazir Hossain, August 2018

Pitzer's Model," The Journal of Chemical Thermodynamics, vol. 19, pp. 185-193,
1987.
[17] V. K. Filippov, A. M. Kalinkin, and S. K. Vasin, "Thermodynamics of Phase
Equilibria of Aqueous (Lithium Sulfate + Alkali-Metal Sulfate) (Alkali Metal:
Na, K, and Rb), and (Sodium Sulfate + Rubidium Sulfate), at 298.15 K using
Pitzer's Model," The Journal of Chemical Thermodynamics, vol. 21, pp. 935-946,
1989.
[18] S. Wang and T. Deng, "(Solid + Liquid) Isothermal Evaporation Phase Equilibria
in the Aqueous Ternary System (Li2SO4+MgSO4+H2O) at T=308.15 K," The
Journal of Chemical Thermodynamics, vol. 40, pp. 1007-1011, 2008.
[19] C. E. Harvie and J. H. Weare, "The Prediction of Mineral Solubilities in Natural
Waters: the Na-K-Mg-Ca-Cl-SO4-H2O System from Zero to High Concentration
at 25 °C," Geochimica et Cosmochimica Acta, vol. 44, pp. 981-997, 1980.
[20] H. Zhou, D. Zeng, H. Han, O. Dong, D. Li, and Y. Yao, "Solubility Isotherm of
the System Li2SO4–K2SO4–MgSO4–H2O at 273.15 K," Journal of Chemical &
Engineering Data, vol. 58, pp. 1692-1696, 2013.
[21] S. L. Clegg and K. S. Pitzer, "Thermodynamics of Multicomponent, Miscible,
Ionic Solutions: Generalized Equations for Symmetrical Electrolytes," The
Journal of Physical Chemistry, vol. 96, pp. 3513-3520, 1992.
[22] S. L. Clegg, K. S. Pitzer, and P. Brimblecombe, "Thermodynamics of
Multicomponent, Miscible, Ionic Solutions. Mixtures including Unsymmetrical
Electrolytes," The Journal of Physical Chemistry, vol. 96, pp. 9470-9479, 1992.
[23] W. Voigt, "Solubility Equilibria in Multicomponent Oceanic Salt Systems from t=
0 to 200 °C. Model Parameterization and Databases," Pure and Applied
Chemistry, vol. 73, pp. 831-844, 2001.
[24] C. C. Chen, H. I. Britt, J. Boston, and L. Evans, "Local Composition Model for
Excess Gibbs Energy of Electrolyte Systems. Part I: Single Solvent, Single
Completely Dissociated Electrolyte Systems," AIChE Journal, vol. 28, pp. 588-
596, 1982.
[25] N. Hossain, S. K. Bhattacharia, and C. C. Chen, "Temperature Dependence of
Interaction Parameters in Electrolyte NRTL Model," AIChE Journal, vol. 62, pp.
1244-1253, 2016.
[26] Y. Song and C.-C. Chen, "Symmetric Electrolyte Nonrandom Two-Liquid
Activity Coefficient Model," Industrial & Engineering Chemistry Research, vol.
48, pp. 7788-7797, 2009.
[27] D. D. Wagman, W. H. Evans, V. B. Parker, R. H. Schumm, and I. Halow, "The
NBS Tables of Chemical Thermodynamic Properties. Selected Values for
Inorganic and C1 and C2 Organic Substances in SI Units," National Standard
Reference Data System 1982.
[28] A. Seidell, Solubilities of Inorganic and Organic Compounds c. 2: D. Van
Nostrand Company, 1919.
[29] R. Robinson, J. M. Wilson, and R. Stokes, "The Activity Coefficients of Lithium,
Sodium and Potassium Sulfate and Sodium Thiosulfate at 25 ° from Isopiestic

116
Texas Tech University, Nazir Hossain, August 2018

Vapor Pressure Measurements," Journal of the American Chemical Society, vol.


63, pp. 1011-1013, 1941.
[30] I. D. Zaytsev and G. G. Aseyev, Properties of Aqueous Solutions of Electrolytes:
CRC press, 1992.
[31] A. N. Campbell and E. M. Kartzmark, "The Systems Li2SO4-K2SO4-H2O and
Li2SO4-Na2SO4-H2O at 25 °C," Canadian Journal of Chemistry, vol. 36, pp. 171-
175, 1958.
[32] S. Tanveer and C.-C. Chen, "Molecular Thermodynamic Model for Aqueous
Na+–K+–Mg2+–Ca2+– SO42− System " In preperation, 2018.

117
Texas Tech University, Nazir Hossain, August 2018

CHAPTER 5
DISSOCIATION BEHAVIOR OF IONIC LIQUIDS IN WATER:
THERMODYNAMIC MODELING AND MOLECULAR DYNAMICS
SIMULATIONS
5.1 Abstract
Ionic liquids are often approximated as either in all molecular form or completely

dissociated in solutions for thermodynamic modeling. In this work, we performed

molecular dynamics simulations for three ionic liquids (ILs): 1-ethyl-

3methylimidazolium ethylsulfate, 1-ethyl-3-methylimidazolium trifluoroacetate, and 1-

ethyl-3-methylimidazolium trifluoromethanelsulfonate in water. Consistent with the

literature experimental data showing these ILs are partially dissociated, the molecular

dynamics simulations show the formation of singlets, doublets, and triplets at very dilute

concentration of ILs in water. Based on these findings, we propose a dissociation

chemistry for ILs and accurately correlate all available vapor pressure and dissociation

extent data over the entire IL concentration range for the three ILs with symmetric

electrolyte Non-Random Two Liquid model.

Keywords: Ionic Liquids; Molecular Dynamics Simulations; Partial Dissociation;

Symmetric Electrolyte Non-Random Two Liquid Model

118
Texas Tech University, Nazir Hossain, August 2018

5.2 Introduction
Considered as “green solvents,” ionic liquids (ILs) have gained a lot of attention

for niche applications, such as CO2 capturing [1-3], separation solvents [4], refrigerants,

reaction media [5, 6], novel electrolytes in batteries and photovoltaic cells [7] for their

negligible vapor pressure [8]. Most of the ILs are very viscous, therefore, it is a common

practice to use solvent such as water to reduce the viscosity [9].

There have been extensive efforts on thermodynamic modeling of IL-containing

mixtures [10, 11], centered around two types of activity coefficient models: correlative

(UNIQUAC [12], NRTL [13], eNRTL [14, 15] etc.) and predictive (UNIFAC [16],

COSMO-RS [17], COSMO-SAC [18], and NRTL-SAC [19]). Correlative models take

advantage of the experimental data to quantify the adjustable interaction parameters,

whereas predictive models make use of molecular structural information such as

functional groups, charge distribution over molecule surface, and molecular segment

information. Among them, Simoni [10] reported an extensive thermodynamic study on

aqueous solutions of 1-ethyl-3methylimidazolium ethylsulfate ([emim][EtSO4]), 1-ethyl-

3-methylimidazolium trifluoroacetate ([emim][TFA]), and 1-ethyl-3-methylimidazolium

trifluoromethanelsulfonate ([emim][TfO]) with the correlative NRTL and eNRTL

models. The chemical structures of the three ILs are shown in Figure 5.1. The ILs were

considered as molecular species in the NRTL model parameterization while they were

treated either as completely dissociated species (𝐶𝐴 → 𝐶 + + 𝐴− ) or with fixed

dissociation extent in the eNRTL model parameterization. The binary interaction

parameters of NRTL and eNRTL models were then regressed from available

experimental binary VLE and activity coefficient at infinitely dilution data. Simoni
119
Texas Tech University, Nazir Hossain, August 2018

considered fixed dissociation extent because their measured dissociation extents were

found to be almost constant throughout the concentration. Recently, Lee and Lin [11]

assumed conventional partial dissociations of ILs (𝐶𝐴 ↔ 𝐶 + + 𝐴− ) for the

[emim][EtSO4] + water binary and the [emim][TfO] + water binary with COSMO-SAC

model [18]. While the model was able to achieve reasonable agreement with the

dissociation extent data in concentrated IL solutions, contrary to the nearly fixed

dissociation extent observed for the IL solutions throughut the concentration [10], the

model suggests a nearly complete dissociation of ILs at the infinite dilution of ILs in

water.

This work aims to reconcile the discrepancy between the experimental IL

dissociation extent data and the complete dissociation of ILs at the infinite dilution in

water suggested by the prior thermodynamic study with the COSMO-SAC model [11].

We first replace COSMO-SAC with the symmetric eNRTL model [20] to confirm the

complete dissociation results at the IL infinite dilution conditions are independent of

choice of activity coefficient models. We then perform molecular dynamics (MD)

simulations to provide molecular insights on the dissociation of ILs in water at the IL

infinite dilution conditions. Based on the evidence of formation of clusters from the MD

simulations, we then propose an alternative dissociation chemistry for the ILs in water to

simultaneously correlate the dissociation extent data [10] and the binary vapor-liquid

equilibrium (VLE) data [10, 21-24] for three ILs, [emim][EtSO4], [emim][TFA], and

[emim][TfO], in water at 298.15 K, 313.15 K, and 333.15 K.

120
Texas Tech University, Nazir Hossain, August 2018

5.3 Experimental Data


Simoni [10] reported dissociation extent data of [emim][EtSO4], [emim][TFA],

and [emim][TfO] in water at 298.15 K, 313.15 K, and 333.15 K by measuring

conductivity, viscosity, and density and applying Nernst-Einstein and Stokes-Einstein

equations. Calvar et al. [21] reported VLE data for the [emim][EtSO4] + water + ethanol

ternary and the corresponding binary systems at multiple temperatures and IL

concentrations. In addition, they reported water activity coefficient data for the

[emim][EtSO4] + water binary. Wang et al. [22] also measured vapor pressures of the

[emim][EtSO4] + water binary at fixed IL compositions. The vapor pressures and water

activity coeficients of the [emim][TFA] + water and [emim][TfO] + water binaries at

various temperatures were reported by Simoni et al. [23]. Orchillés et al. [24] reported

isobaric VLE data of [emim][TfO] + water, [emim][TfO] + propanol, and [emim][TfO] +

water + propanol at 100 kPa. Table 5.1 summarizes available experimental dissociation

extent and VLE data for the three IL-water binary systems in the order of [emim][EtSO4],

[emim][TFA], and [emim][TfO].

5.4 Symmetric eNRTL Model with Conventional Dissociation Chemistry


Assuming the conventional partial dissociation chemistry, we first apply the

symmetric eNRTL model [20] to the three IL-water binaries and compare the

thermodynamic modeling results against those obtained by Lee and Lin [11] with

COSMO-SAC to confirm the complete dissociation results at the IL infinite dilution

conditions are independent of choice of activity coefficient models. The symmetric

eNRTL model provides a thermodynamic framework for electrolytes with symmetric

reference states [20] in the sense that pure liquid state and pure fused salt state are chosen
121
Texas Tech University, Nazir Hossain, August 2018

as the reference states for solvents and electrolytes respectively. This is in contrast to the

conventional electrolyte thermodynamic treatment, i.e., unsymmetric convention, with

which the pure liquid state is chosen as the reference state for solvents and aqueous phase

infinite dilution state is chosen as the refernce state for electrolytes. A particular

advantageous property of the eNRTL model is that the model requires only two

asymmetric binary interaction parameters (𝜏𝑖𝑗 ≠ 𝜏𝑗𝑖 ) to be adjusted for each 𝑖 − 𝑗

interacting pair. The readers are referred to the abundant literature for the exact model

formulation of symmetric eNRTL model and its application for conventional electrolytes

[14, 20, 25-27].

Similar to the COSMO-SAC modeling study, we consider that ILs (CA) undergo

partially dissociation in water by the following conventional dissociation chemistry

(Chemistry I):
𝐾𝐼
𝐶𝐴 ↔ 𝐶 + + 𝐴− (I)

where 𝐾𝐼 is the chemical equilibrium constant. The number of moles of the undissociated

species (𝐶𝐴) and those of the dissociated species (𝐶 + and 𝐴− ) would be


0
𝑛𝐶𝐴 = (1 − 𝜉) 𝑛𝐶𝐴 (5.1)
0
𝑛𝐶 + = 𝑛𝐴− = 𝜉 𝑛𝐶𝐴 (5.2)
0
Where, 𝜉 is the degree of dissociation extent and 𝑛𝐶𝐴 is the total number of moles of IL

prior to the dissociation. The chemical equilibrium constant can be expressed in terms of

the species mole fractions 𝑥𝑖 and the species activity coefficients 𝛾𝑖 as follows:
𝑥𝐶+ 𝛾𝐶+ 𝑥𝐴− 𝛾𝐴−
𝐾𝐼 = (5.3)
𝑥𝐶𝐴 𝛾𝐶𝐴

where the activity coefficients are calculated with the symmetric eNRTL model.

122
Texas Tech University, Nazir Hossain, August 2018

The chemical equilibrium constant, 𝐾𝐼 , and the eNRTL binary interaction

parameters are adjusted to match the dissociation extent data [10] and the VLE data [10,

21-24].

In the regression, the temperature data is treated as error-free and the standard

deviation of the IL mole fraction data and that of the dissociation extent data were set to

0.01 and 0.02, respectively.

For the isobaric [emim][EtSO4] + water VLE data at 0.1013 MPa [21], the

standard deviation in ILs mole fraction was set to 0.01 and the standard deviation in

temperature was set to 0.1 K in regression. For the water activity coefficient data [21], the

temperature is error-free and the standard deviation for the IL mole fraction is 0.01 and

the standard deviation for the water activity coefficient is set to 2%. For the TPx VLE

data set of [emim][EtSO4] + water [22], the standard deviations in temperature and

pressure were set to 0.1 K and 3%, respectively. The compositions are assumed to be

error-free.

For the isothermal [emim][TFA] + water VLE data and the isothermal

[emim][TfO] + water binary VLE data [23], the standard deviations in IL mole fraction

and pressure were set to 0.01 and 3%, respectively. For the water activity coefficient data

for the [emim][TFA] + water and [emim][TfO] + water binary binary [23], the

temperature is error-free and the standard deviation for the IL mole fraction is 0.01 and

the standard deviation for the water activity coefficient is set to 2%. However, some of

the water activity coefficient data for the [emim][TfO] + water binary [23] are not used in

regression because those activity coefficients are more than unity which indicates the
123
Texas Tech University, Nazir Hossain, August 2018

formation of two liquid phases. The standard deviations in temperature was set to 0.1 K,

the pressure error-free, and the standard deviation for the IL mole fraction is set to 0.01

for the isobaric [emim][TfO] + water binary VLE data [24]. ASPEN V8.4 [28] was used

for regression and prediction. Additionally, mean relative deviation (MRD) between the

model results and the experimental data are reported to illustrate the goodness of fit.

1 𝐸𝑠𝑡−𝐸𝑥𝑝
𝑀𝑅𝐷 % = 𝑘 ∑𝑘𝑖=1 ( ) × 100 (5.4)
𝐸𝑥𝑝

Table 5.1 shows the MRD of dissociation extent data and VLE data by

considering Chemistry I. The MRD values indicate that eNRTL model correlates all VLE

data very well

Table 5.2 and Table 5.3 summarize the identified eNRTL binary interaction

parameters and the chemical equilibrium constant with Chemistry I, respectively. Figure

5.2 shows the comparison of the model predictions and the experimental data of the

dissociation extent for the three ILs in water at 298.15 K. The dissociation extents are

correlated very well throughout the concentration. However, contrary to the trend of the

experimental data [10], the eNRTL model results suggest complete dissociation of ILs at

infinitely dilution of ILs in water, in line with the COSMO-SAC model predictions [11].

Although not shown in the text, all three ILs show similar behavior at the other two

temperatures (313.15 K and 333.15 K). The failure to represent the dissociation extent

data at the infinite dilution conditions suggests that further study on the IL dissociation

chemistry is needed.

124
Texas Tech University, Nazir Hossain, August 2018

5.5 Molecular Dynamics Simulation

5.5.1 Prior Works


There are numerous studies of molecular dynamics simulations of ILs available in

the literature, especially focusing on the dynamic properties i.e. conductivity, diffusivity,

and viscosity [29-35]. The degree of dissociation of ILs can be correlated from these

dynamic properties with Nernst-Einstein and Stokes-Einstein equation [10]. The degree

of associations of ILs, so called “ionicity”, which can be found from the ratio of

experimentally measured conductivity and ideal conductivity from Nernst-Einstein

relation, vary from 0.01-0.9 [36, 37]. ILs show lower conductivities than the conventional

electrolytes due to the formation of ion pairs or larger aggregates [38]. Zhao et al.

suggested that the electrostatic drag causes the lowering the conductivity of ILs [39].

Zhang and Maginn showed that “ion pair (IP)” and “ion cage (IC)” lifetimes are inversely

correlated with diffusivity and conductivity for 29 ILs [40]. Ion-pairing of ILs can also be

found in dilute solution of ILs, both in experiments [41, 42] and in molecular dynamics

simulations [43]. Bernardes et al. [44] studied [emim][EtSO4] in water at full

concentration range and found that the ions form smaller clusters from high concentration

to low concentration. At very dilute IL concentration, 60% ions are found to be isolated

[44]. Further study is needed to be done on cluster mechanism of ILs at very dilute

concentration to explain the dissociation behavior.

5.5.2 Simulation Methods


To decipher the IL dissociation chemistry, we perform MD simulations and study

the ion pairing mechanism of the IL+ water binaries. The summary of the prepared

simulated systems is shown in Table 5.4. Around 2000 IL molecules are considered for
125
Texas Tech University, Nazir Hossain, August 2018

pure ILs while around 40,000 molecules are considered for ILs in infinite dilution of

aqueous IL solutions. The interatomic interactions of all particles except water were

taken into account by the general AMBER force field (GAFF) [45, 46] parameters, and

TIP3P model [47, 48] was used to calculate interactions between water molecules. The

Lennard-Jones interaction parameters between water and ILs were estimated by using

Lorenz-Berthelot mixing rules [49]. The bond length of the water molecules was

constrained by the SHAKE algorithm [50]. The partial charges of the atoms were

determined by the AM1-BCC method [51, 52]. The partial charges of [emim][EtSO4],

[emim][TFA], and [emim][TfO] were rescaled to ± 0.80, ± 0.90, and ± 0.80 respectively

to take into account for the polarizability of the ILs [39, 40, 53]. The potential energy due

to electrostatic and van der Waals interactions were determined within the cut off

distance of 12 Å. The long range potential energy part was calculated by the tail

correction and particle-particle particle-mesh (PPPM) algorithm [54]. Temperature and

pressure were controlled by using Nosé−Hoover thermostat and barostat, respectively

[55, 56]. The LAMMPS [57] package was used to perform all MD simulations having

time step of 1 fs.

Following the simulation method of Zhang et al. [58], we first prepare models of

single molecules of ILs and water separately. The prepared ILs and water molecules were

then replicated and mixed to prepare their binary mixture at different concentrations. All

simulations were carried out using constant NPT (constant number of particle, pressure,

and temperature) molecular dynamics simulation at 298.15 K and 0.1013 MPa.

126
Texas Tech University, Nazir Hossain, August 2018

The force field is validated by comparing simulated density of the pure ILs with

experimental data [44, 59]. The simulated density is compared with experimental density

in Table 5.5. It can be seen that the simulated density of the pure ILs is within the range

of experimental density.

5.5.3 Structural Properties


The structural properties can be estimated by the radial distribution function (RDF),

g(r), of the systems. The RDF is the probability of finding an atom/molecule at distance r

from another atom/molecule. The RDF was calculated at 298.15 K for all systems by

time averaging over last 10 ns of total 50 ns simulation times. The center of mass RDF

between cation and anion of each system is shown in Figure 5.3. The first peak between

cation and anion is observed at a distance of about 5 Å for all systems. The intensity of

cation-anion RDF first peak is higher than that of the either cation-cation or anion-anion

pairs. Therefore, we might conclude that there is a stronger interaction between the

cation-anion pair than other pairs.

To see the interactions between the specific atoms within the cations and anions, we

also show the RDF between the most acidic hydrogen atom of the cation and the oxygen

atom of the anion in Figure 5.4. All ionic liquids show a strong first peak at a distance of

about 2.5 Å (see Figure 5.4). From this observation, it is clear that the most acidic

hydrogen atom of the cation and the oxygen atom of the anion interacts strongly to form

clusters of ionic species.

127
Texas Tech University, Nazir Hossain, August 2018

5.5.4 Clustering of Ionic Liquids


The systems were relaxed for 50 ns, and the clustering analysis was carried out

for last 10 ns. The cut off distance for clustering calculation was set to the first minima of

radial distribution function [44, 60, 61]. The probability of finding a cluster of size n is

defined as [44, 60, 61]

𝑛 ∑𝑁
𝑖=1 𝑆𝑛 (𝑖)
𝑃(𝑛) = (5.5)
𝑁𝐼

where P(n), N, Sn, and I represent the probability of finding cluster with size n, the total

number of snapshots taken from the simulation, the number of cluster of size n, and the

total number of ions (cation, C and anion, A) in the systems, respectively.

Figure 5.5 shows the probability of finding cluster of size n at 0.5 mol% of ILs in

water. The ions form smaller clusters for [emim][EtSO4] and [emim][TFA]. Bernardes et

al. [44] also reported that at 0.4 mol% of [emim][EtSO4] in water forms smaller

aggregates. We found that [emim][TfO] forms higher aggregates which indicates the

formation of two liquids at 0.5 mol% concentration. For 0.5 mol% ILs in water, most of

the aggregates are smaller than 20 ions, except [emim][TfO]. We have analyzed clusters

containing 1, 2, 3, 4, and 5 ions in more detail. We quantified the amount of species in

400 ionic species (200 IL molecules) in a given cluster size. Table 5.6 shows the number

of cluster elements in the cluster sizes up to 5.

We calculated the degree of dissociation, up to cluster size 5, for these three ILs

by counting the charged species (positive/negative) in the clusters by following:


1
(∑ 𝑧𝑖 (𝐶𝑙𝑢𝑠𝑡𝑒𝑟)𝑖 )
𝜉 = 𝑇𝑜𝑡𝑎𝑙 𝐼𝐿 𝑚𝑜𝑙𝑒𝑐𝑢𝑙𝑒𝑠
2
(5.6)
𝑝𝑟𝑖𝑜𝑟 𝑡𝑜 𝑑𝑖𝑠𝑠𝑜𝑐𝑖𝑎𝑡𝑖𝑜𝑛

128
Texas Tech University, Nazir Hossain, August 2018

Here, clusters are singlets (S: C and A), doublets (D: CA, CC, and AA), and

triplets (T: CAC, ACA, CCC, AAA) and so on. And 𝑧𝑖 is the charge number of the

particular cluster. For example, in singlets (𝑧𝐶 = 𝑧𝐴 = 1), in doublets (𝑧𝐶𝐴 = 0, 𝑧𝐶𝐶 =

𝑧𝐴𝐴 = 2), in triplets (𝑧𝐶𝐴𝐶 = 𝑧𝐴𝐶𝐴 = 1, 𝑧𝐶𝐶𝐶 = 𝑧𝐴𝐴𝐴 = 3), and so on.

Figure 5.6 shows the dissociation extent of ILs up to cluster size 5. The

dissociation extents have reaches a steady value after cluster size 3. For the simplicity in

thermodynamic modeling of aqueous ILs, which will be discussed in the later section, we

will assume that the ILs form only singlet (S: C and A), doublets (D: CA, CC, and AA),

and triplets (T: CAC, ACA, CCC, AAA). Table 5.7 shows the dissociation extents of

[emim][EtSO4], [emim][TFA], and [emim][TfO] from molecular dynamics simulations

up to cluster size 3. The dissociation extents are within the range of experimental

dissociation extent data [10]. It should be noted that the dissociation extents for

[emim][TfO] are lower than other two ILs because molecular dynamics simulation shows

a phase separation and most of the species form bigger clusters.

5.6 Symmetric eNRTL Model with New Dissociation Chemistry


The above MD simulations provide the evidence of the formation of singlets,

doublets, and triplets, mostly. Within the doublets, the numbers of double charged

species, i.e. CC and AA, are insignificant in comparison to that of neutral species, i.e.,

CA (See Table 5.6). Similarly, for the triplets, the numbers of multiple charged species,

i.e., CCC and AAA, are very small when compared to those of single charged species,

i.e., CAC and ACA. The simulation results depict a picture in which the ILs dissolve in

solution as singlets C and A, doublet CA, and triplets CAC and ACA. In other words, the

129
Texas Tech University, Nazir Hossain, August 2018

conventional partial dissociation chemistry (Chemistry I) is not supported by the

simulation results. Rather, the simulation results suggest the following chemistry for ILs

(Chemistry II):
𝐾𝐼𝐼
2 𝐶𝐴 (𝐷) ↔ 𝑇 +/− + 𝑆 −/+ (II)

where D is the doublet (CA), T the triplet (CAC or ACA), S the singlet (C or A),

and 𝐾𝐼𝐼 the chemical equilibrium constant.

The degree of dissociation (𝜉) can be calculated as follows:


𝑛 −/+ 𝑛 −/+
𝜉=𝑛 𝑆
=𝑛 𝑇
(5.7)
𝐷 + 𝑛𝑆−/+ 𝐷 + 𝑛𝑇−/+

Table 5.8 shows the regressed binary interaction parameters of eNRTL model for

the [emim][EtSO4] + water, [emim][TFA] + water, and [emim][TfO] + water binary

systems by considering Chemistry II. Table 5.9 shows the corresponding chemical

equilibrium constants for the partial dissociation of these three ILs based on Chemistry II.

Figure 5.7, 5.8, and 5.9 present the comparison of the model correlation results and the

experimental data of the dissociation extent [10] of the three ILs in water at 298.15 K,

313.15 K, and 333.15 K, respectively. The dissociation extents are correlated very well

throughout the concentration at the three temperatures. Shown in Table 5.1, the MRDs

for [emim][EtSO4], [emim][TFA], and [emim][TfO] are 6.2, 5.9, and 6.9%, respectively,

which are significantly better than the results with Chemistry I. In short, by considering

Chemistry II, eNRTL successfully correlates the dissociation extent data for all three ILs,

including the dilute IL regions.

We also compared the VLE data of ILs + water binaries with eNRTL model

results by considering Chemistry II. Figure 5.10 and Figure 5.11 represent the
130
Texas Tech University, Nazir Hossain, August 2018

comparisons of the isobaric VLE data and water activity coefficients [21] of

[emim][EtSO4] + water binary at 0.101 MPa with eNRTL model results. The model

results are in excellent agreement with the experimental data. Figure 5.12 compares VLE

data [22] at fixed compositions of [emim][EtSO4] + water binary with eNRTL model

predictions. It is clear that the model correlates the VLE data very well.

Figure 5.13 and Figure 5.14 depict the comparisons of isothermal VLE data and

corresponding water activity coefficient of [emim][TFA] + water binary [23] with

eNRTL model results, respectively. In both cases, model results are in good agreement

with experimental data.

Figure 5.15 represents the comparisons of isothermal VLE data of [emim][TfO] +

water binary [23] with eNRTL model results. The model predictions are in line with

experimental data. Figure 5.16 shows the comparison of isobaric VLE data [24] with

eNRTL model predictions. It has been seen that the model results are in excellent

agreement with data. The water activity of [emim][TfO] + water binary at various

temperatures are shown in the Figure 5.17. The agreement between the model results and

experimental data are reasonably well. Although the experimental data suggest that there

is a possibility of forming two liquid phases for [emim][TfO] + water binary in some

compositions and temperatures, the eNRTL model does not predict any two liquid

phases.

Overall, the predictions of VLE of ILs + water binary with eNRTL model are

similar regardless of choosing dissociation chemistry of ILs. However, the eNRTL model

131
Texas Tech University, Nazir Hossain, August 2018

with Chemistry II is able to represent the dissociation extent of ILs in water in the entire

concentration ranges.

5.7 Conclusions
Like all electrolytes, properly accounting for the dissociation chemistry of ionic

liquids, whether in pure or in water, is essential. COSMO-SAC and eNRTL model

predict complete dissociation of ILs in water at infinitely dilute with the conventional

partial dissociation chemistry. However, molecular dynamics simulation shows that ILs

form singlets, doublets, and triplets, primarily. We proposed a new dissociation chemistry

for ILs in water in symmetric eNRTL model, in accordance with molecular dynamics

simulation. The symmetric eNRTL model, with new dissociation chemistry, are able to

correlate dissociation extent and binary vapor liquid equilibrium data with reasonable

accuracy over the entire temperature and concentration range. The clustering of the Ionic

Liquids in water will be further studied in quantum mechanical approach.

132
Texas Tech University, Nazir Hossain, August 2018

Tables
Table 5.1 Literature data and corresponding MRD (%).

Mole No. of MRD (%)


Temperature Pressure
Data Type fraction of data Source Chemistry Chemistry
(K) (MPa)
ILs points I II
[emim][EtSO4] + H2O
Dissociation
298.15-333.15 0.1 0.00-1.00 51 Simoni [10] 17.1 6.2
extent
Isobaric VLE 373.15-410.58 0.101 0.00-0.42 22 Calvar [21] 0.1 a 0.1 a
Activity
coefficient of 373.15-410.58 0.101 0.00-0.42 22 Calvar [21] 1.5 2.0
water
VLE at fixed
316.93-369.29 0.008-0.070 0.009-0.151 32 Wang et al. [22] 1.3 b 0.9 b
composition
[emim][TFA] + H2O
Dissociation
298.15-333.15 0.1 0.00-1.00 30 Simoni [10] 21.0 5.9
extent
Isothermal VLE 328.2-368.2 0.006-0.088 0-0.088 37 Simoni et al. [23] 1.6 b 2.0 b
Activity
coefficient of 328.2-368.2 0.006-0.088 0-0.088 37 Simoni et al. [23] 1.5 2.0
water
[emim][TfO] + H2O
Dissociation
298.15-333.15 0.1 0.00-1.00 51 Simoni [10] 10.3 6.9
extent
Isothermal VLE 323.3-363.3 0.006-0.0693 0.01-0.5 27 Simoni et al. [23] 4.4 b 4.2 b

133
Texas Tech University, Nazir Hossain, August 2018

Table 5.2 Continued

Isobaric VLE 372.81-391.10 0.1 0.00-0.42 31 Orchillés et al. [24] 0.03 a 0.05 a
Activity
**
coefficient of 323.3-363.3 0.006-0.0693 0.01-0.5 27 (11) Simoni et al. [23] 4.5 4.3
water

a
MRD for temperature
b
MRD for pressure
*Excess enthalpy data are not included in the regression.
**
For water activity coefficient of water of [emim][TfO] + H2O binary, 11 data points are included in the regression since other data
points are more than unity.

134
Texas Tech University, Nazir Hossain, August 2018

Table 5.3 Binary interaction parameters of eNRTL model with chemistry I for
[emim][EtSO4], [emim][TFA], and [emim][TfO] in water.

Component i Component j 𝐶𝑖𝑗 𝐷𝑖𝑗 a


𝜏𝑖𝑗 (298.15 K)

H2O ([emim]+ [EtSO4]-) -1.819 ± 4.621 547.990 ± 150.628 0.019 ± 6.846

([emim]+ [EtSO4]-) H2O -5.527 ± 5.253 1617.531 ± 1897.150 -0.101 ± 8.251


H2O [emim][EtSO4] -2.997 ± 63.203 1632.522 ± 6192.786 2.478 ± 92.053
[emim][EtSO4] H2O -1.010 ± 19.804 -372.043 ± 6192.786 -2.258 ± 28.699
H2O ([emim]+ [TFA]-) -7.177 ± 2.171 1970.900 ± 715.553 -0.566 ± 4.571
([emim]+ [TFA]-) H2O 4.512 ± 2.421 -1570.306 ± 823.406 -0.755 ± 5.182
H2O [emim][TFA] 10.452 ± 8.040 -2097.547 ± 2799.743 3.417 ± 17.430
[emim][TFA] H2O -2.131 ± 1.250 -275.491 ± 439.005 -3.055 ± 2.722

H2O ([emim]+ [TfO]-) -1.340 ± 2.873 2795.801 ± 963.657 8.037 ± 4.324

([emim]+ [TfO]-) H2O -1.771 ± 1.286 -473.542 ± 424.568 -3.359 ± 1.919


H2O [emim][TfO] -15.683 ± 4.000 6107.290 ± 1286.436 4.801± 5.883

[emim][TfO] H2O 6.999 ± 2.863 -2783.223 ± 897.635 -2.336 ± 4.155

𝐷𝑖𝑗
a
𝜏𝑖𝑗 = 𝐶𝑖𝑗 + 𝑇

135
Texas Tech University, Nazir Hossain, August 2018

Table 5.4 Equilibrium constants of Ionic Liquids (ILs) for chemistry I.

IL a b a
𝑙𝑛𝐾𝐼 (298.15 K)

[emim][EtSO4] -6.229 ± 2.226 1319.880 ± 705.413 -1.802 ± 3.249

[emim][TFA] -3.516 ± 1.387 334.845 ± 433.222 -2.393 ± 2.840

[emim][TfO] -6.434 ± 1.022 1292.635 ± 318.137 -2.099 ± 1.477

𝑏
a
𝑙𝑛 𝐾𝐼 = 𝑎 + 𝑇

136
Texas Tech University, Nazir Hossain, August 2018

Table 5.5 The simulation box composition of the binary mixtures of water and ionic
liquids.

[emim][EtSO4] [emim][TFA] [emim][TfO]

𝑋𝐼𝐿𝑠 𝑁𝐼𝐿𝑠 𝑁𝐻2 𝑂 𝑁𝐼𝐿𝑠 𝑁𝐻2 𝑂 𝑁𝐼𝐿𝑠 𝑁𝐻2 𝑂

1.000 2100 0 2300 0 2200 0

0.005 200 39200 200 39200 200 39200

𝑋𝐼𝐿𝑠 , 𝑁𝐼𝐿𝑠 , and 𝑁𝐻2 𝑂 are the mole fraction of ionic liquid, the number of ion pairs, and the
number of water molecules, respectively.

137
Texas Tech University, Nazir Hossain, August 2018

Table 5.6 The simulated and experimental density (g/cc) of the pure ionic liquids at
298.15 K and atmospheric pressure.

[emim][EtSO4] [emim][TFA] [emim][TfO]

Simulation 1.2427 ± 0.0002 1.2935 ± 0.0003 1.3807 ± 0.0002

Experiment 1.2374 [44] 1.2908 [44] 1.3836 [59]

138
Texas Tech University, Nazir Hossain, August 2018

Table 5.7 The number of cluster elements in clusters size up to 5.

n=1 n=2 n=3

C A CC CA AA ACA AAA CAC CCC

[emim][EtSO4] 47.96 ± 0.87 37.53 ± 0.80 2.03 ± 0.13 24.83 ± 1.21 2.04 ± 0.03 7.36 ± 0.74 0.12 ± 0.01 6.44 ± 0.56 0.09 ± 0.02

[emim][TFA] 75.26 ± 2.26 72.95 ± 1.71 4.09 ± 0.03 40.18 ± 0.38 3.98 ± 0.08 11.08 ± 0.08 0.28 ± 0.01 9.09 ± 0.03 0.19 ± 0.01

[emim][TfO] 22.23 ± 1.20 15.82 ± 1.20 0.26 ± 0.04 3.92 ± 0.43 0.33 ± 0.03 0.47 ± 0.05 0.01 ± 0.00 0.33 ± 0.08 0.00 ± 0.00

n=4 n=5

AAAA CAAA CCAA CCCA AAAAA CAAAA CCAAA CCCAA CCCCA CCCCC

0.08 ± 0.07 1.72 ± 0.41 3.63 ± 0.92 0.00 ± 0.00 0.00 ± 0.00 0.22 ± 0.02 2.19 ± 0.26 1.76 ± 0.27 0.16 ± 0.01 0.00 ± 0.00

0.02 ± 0.01 1.86 ± 0.03 6.23 ± 0.09 0.00 ± 0.00 0.00 ± 0.00 0.27 ± 0.03 2.36 ± 0.03 2.21 ± 0.06 0.15 ± 0.02 0.00 ± 0.00

0.00 ± 0.00 0.03 ± 0.01 0.09 ± 0.02 0.00 ± 0.00 0.00 ± 0.00 0.00 ± 0.00 0.03 ± 0.01 0.02 ± 0.01 0.00 ± 0.00 0.00 ± 0.00

139
Texas Tech University, Nazir Hossain, August 2018

Table 5.8 Dissociation extent (ξ) from molecular dynamics simulation of 0.5 mol% ILs
in water up to cluster size 3.

IL 𝜉

[emim][EtSO4] 0.2702 ± 0.0044

[emim][TFA] 0.4648 ± 0.0100

[emim][TfO] 0.1001 ± 0.0043

140
Texas Tech University, Nazir Hossain, August 2018

Table 5.9 Binary interaction parameters of eNRTL model with Chemistry II for
[emim][EtSO4], [emim][TFA], and [emim][TfO] in water.

Component Component
𝐶𝑖𝑗 𝐷𝑖𝑗 a
𝜏𝑖𝑗 (298.15 K)
i j

[emim][EtSO4]

H2O (T+/T- S-/S+) 0.000 ± 2.231 484.715 ± 775.110 1.629 ± 3.426

(T+/T- S-/S+) H2O -3.418 ± 0.125 -0.043 ± 9.435 -3.418 ± 0.129


H2O D (CA) 4.810 ± 20.246 -968.463 ± 5898.967 1.561 ± 28.308
D (CA) H2O -1.745 ± 5.601 -288.867 ± 1634.876 -2.714 ± 7.838

[emim][TFA]

H2O (T+/T- S-/S+) -12.329 ± 2.168 4068.375 ± 719.903 1.317 ± 4.583


(T+/T- S-/S+) H2O 2.381 ± 1.073 -1689.266 ± 369.437 -3.285 ± 2.312
H2O D (CA) 13.606 ± 6.104 -3331.875 ± 2088.875 2.433 ± 13.110
D(CA) H2O -3.582 ± 1.245 242.600 ± 429.547 -2.768 ± 2.686

[emim][TfO]

H2O (T+/T- S-/S+) -2.239 ± 3.863 2998.398 ± 1325.597 7.817 ± 5.890

(T+/T- S-/S+) H2O -1.701 ± 1.398 -654.355 ± 467.640 -3.896 ± 2.101


H2O D (CA) -16.202 ± 3.355 5872.432 ± 1170.377 3.494 ± 5.164

D (CA) H2O 8.896 ± 2.547 -3231.754 ± 746.504 -1.943 ± 3.572

𝐷𝑖𝑗
a
𝜏𝑖𝑗 = 𝐶𝑖𝑗 + 𝑇

141
Texas Tech University, Nazir Hossain, August 2018

Table 5.10 Equilibrium constants of Ionic Liquids (ILs) for chemistry II

IL a b a
𝑙𝑛𝐾𝐼𝐼 (298.15 K)
1317.612 ±
[emim][EtSO4] -5.680 ± 1.357 -1.261 ± 1.968
424.823

[emim][TFA] -3.141 ± 1.561 332.658 ± 488.022 -2.025 ± 3.198

1323.849 ±
[emim][TfO] -6.113 ± 1.552 -1.674 ± 2.246
483.891

𝑏
a
𝑙𝑛𝐾𝐼𝐼 = 𝑎 + 𝑇

142
Texas Tech University, Nazir Hossain, August 2018

Figures

Figure 5.1 The chemical structure of (a) [emim][EtSO4], (b) [emim][TFA], and (c)
[emim][TfO].

143
Texas Tech University, Nazir Hossain, August 2018

Figure 5.2 Degree of dissociation of ILs in water: comparison of experimental data in


symbols with eNRTL model results (Chemistry I) in lines at 298.15 K: ( ), ( )
[emim][EtSO4]; ( ), ( ) [emim][TFA]; and ( ), ( ) [emim][TfO]

144
Texas Tech University, Nazir Hossain, August 2018

Figure 5.3 The center of mass coordinate radial distribution function (g(r)) of ionic
liquids studied. (a), (b), and (c) represents 0.5 mol% [emim][EtSO4], [emim][TFA], and
[emim][TfO] in water, respectively. The legends of the plot are given in the following
way: cation-anion (red line), anion-anion (green line), and cation-cation (black line),
respectively.

145
Texas Tech University, Nazir Hossain, August 2018

Figure 5.4 The RDF between the most acidic hydrogen atom of the cation and the
oxygen atom of the anion: (a) [emim][EtSO4], (b) [emim][TFA], and (c) [emim][TfO].
The red and green curves represent 0.5 mol% ILs in water and pure ILs, respectively.

146
Texas Tech University, Nazir Hossain, August 2018

Figure 5.5 Cluster size distribution of ILs containing 99.5 mol% water: (a)
[emim][EtSO4], (b) [emim][TFA], and (c) [emim][TfO]. The inset of Figure 5.5(a)
shows the cluster size distribution at the low range of cluster size.

147
Texas Tech University, Nazir Hossain, August 2018

Figure 5.6 Degree of dissociation of ILs from molecular dynamics simulation up to


cluster size 5 at 298.15 K

148
Texas Tech University, Nazir Hossain, August 2018

Figure 5.7 Degree of dissociation of [emim][EtSO4] in water: comparison of


experimental data in symbols with eNRTL model results (Chemistry II) in lines at
different temperatures: ( ), ( ) 298.15 K; ( ), ( ) 313.15 K; and (
), ( ) 333.15 K.

149
Texas Tech University, Nazir Hossain, August 2018

Figure 5.8 Degree of dissociation of [emim][TFA] in water: comparison of experimental


data in symbols with eNRTL model results (Chemistry II) in lines at different
temperatures: ( ), ( ) 298.15 K; ( ), ( ) 313.15 K; and ( ), (
) 333.15 K.

150
Texas Tech University, Nazir Hossain, August 2018

Figure 5.9 Degree of dissociation of [emim][TfO] in water: comparison of experimental


data in symbols with eNRTL model results (Chemistry II) in lines at different
temperatures: ( ), ( ) 298.15 K; ( ), ( ) 313.15 K; and ( ), (
) 333.15 K.

151
Texas Tech University, Nazir Hossain, August 2018

Figure 5.10 Isobaric VLE of [emim][EtSO4] + H2O binary at 0.1013 MPa: comparison of
experimental data [21] shown in symbols with eNRTL model results (Chemistry II): (
), ( ) eNRTL model.

152
Texas Tech University, Nazir Hossain, August 2018

Figure 5.11 Activity coefficient of water of [emim][EtSO4] + H2O binary at 0.1013 MPa:
comparison of experimental data shown in symbols with eNRTL model results
(Chemistry II): ( ) experiment [21], ( ) eNRTL model.

153
Texas Tech University, Nazir Hossain, August 2018

Figure 5.12 VLE of [emim][EtSO4] + H2O binary at fixed mole fractions of


[emim][EtSO4]: comparison of experimental data [22] shown in symbols with eNRTL
model results (Chemistry II) shown as lines: ( ), ( ) 0.0085; ( ), ( )
0.0709; ( ), ( ) 0.1511.

154
Texas Tech University, Nazir Hossain, August 2018

Figure 5.13 Isothermal VLE of [emim][TFA] + H2O binary at different temperatures:


comparison of experimental data [23] shown in symbols with eNRTL model results
(Chemistry II) shown as dashed lines: ( ), ( ) 328.3 K; ( ), ( ) 348.2
K; ( ), ( ) 358.2 K ; ( ), ( ) 368.2 K.

155
Texas Tech University, Nazir Hossain, August 2018

Figure 5.14 Activity coefficient of water of [emim][TFA] + H2O binary at different


temperatures: comparison of experimental data [23] shown in symbols with eNRTL
model results (Chemistry II) shown as dashed lines: ( ), ( ) 328.3 K; ( ), (
) 348.2 K; ( ), ( ) 358.2 K ; ( ), ( ) 368.2 K.

156
Texas Tech University, Nazir Hossain, August 2018

Figure 5.15 Isothermal VLE of [emim][TfO] + H2O binary at different temperatures:


comparison of experimental data [23] shown in symbols with eNRTL model results
(Chemistry II) shown as dotted lines: ( ), ( ) 323.3 K; ( ), ( ) 333.3
K; ( ), ( ) 343.3 K ; ( ), ( ) 353.3 K; ( X ), ( ) 363.3 K.

157
Texas Tech University, Nazir Hossain, August 2018

Figure 5.16 Isobaric VLE of [emim][TfO] + H2O binary at 0.1013 MPa: comparison of
experimental data [24] shown in symbols with eNRTL model results (Chemistry II): (
), ( ) eNRTL model.

158
Texas Tech University, Nazir Hossain, August 2018

Figure 5.17 Activity coefficient of water of [emim][TfO] + H2O binary at different


temperatures: comparison of experimental data [23] shown in symbols with eNRTL
model results (Chemistry II) shown as dotted lines: ( ), ( ) 323.3 K; ( ), (
) 333.3 K; ( ), ( ) 343.3 K ; ( ), ( ) 353.3 K; ( X ), ( )
363.3 K.

159
Texas Tech University, Nazir Hossain, August 2018

References

[1] E. D. Bates, R. D. Mayton, I. Ntai, and J. H. Davis, "CO2 Capture by a Task-


Specific Ionic Liquid," Journal of the American Chemical Society, vol. 124, pp.
926-927, 2002.
[2] M. W. Arshad, "CO2 Capture Using Ionic Liquids," Technical University of
Denmark, DTU, DK-2800 Kgs. Lyngby, Denmark, 2009.
[3] M. Hasib-ur-Rahman, M. Siaj, and F. Larachi, "Ionic Liquids for CO2 Capture—
Development and Progress," Chemical Engineering and Processing: Process
Intensification, vol. 49, pp. 313-322, 2010.
[4] A. Chapeaux, L. D. Simoni, T. S. Ronan, M. A. Stadtherr, and J. F. Brennecke,
"Extraction of Alcohols from Water with 1-hexyl-3-methylimidazolium
bis(trifluoromethylsulfonyl)imide," Green Chemistry, vol. 10, pp. 1301-1306,
2008.
[5] K. Seddon, "Room‐Temperature Ionic Liquids: Neoteric Solvents for Clean
Catalysis," ChemInform, vol. 28, 1997.
[6] C. Reichardt, "Solvents and Solvent Effects:  An Introduction," Organic Process
Research & Development, vol. 11, pp. 105-113, 2007.
[7] K. Hayamizu, Y. Aihara, H. Nakagawa, T. Nukuda, and W. S. Price, "Ionic
Conduction and Ion Diffusion in Binary Room-Temperature Ionic Liquids
Composed of [emim][BF4] and LiBF4," The Journal of Physical Chemistry B, vol.
108, pp. 19527-19532, 2004.
[8] K. R. Seddon, "Ionic Liquids for Clean Technology," Journal of Chemical
Technology and Biotechnology, vol. 68, pp. 351-356, 1997.
[9] D. Constantinescu, K. Schaber, F. Agel, M. H. Klingele, and T. J. S. Schubert,
"Viscosities, Vapor Pressures, and Excess Enthalpies of Choline Lactate + Water,
Choline Glycolate + Water, and Choline Methanesulfonate + Water Systems,"
Journal of Chemical & Engineering Data, vol. 52, pp. 1280-1285, 2007.
[10] L. D. Simoni, Predictive modeling of fluid phase equilibria for systems containing
ionic liquids: University of Notre Dame, 2010.
[11] B.-S. Lee and S.-T. Lin, "A Priori Prediction of Dissociation Phenomena and
Phase Behaviors of Ionic Liquids," Industrial & Engineering Chemistry Research,
vol. 54, pp. 9005-9012, 2015.
[12] D. S. Abrams and J. M. Prausnitz, "Statistical Thermodynamics of Liquid
Mixtures: A New Expression for the Excess Gibbs Energy of Partly or
Completely Miscible Systems," AIChE Journal, vol. 21, pp. 116-128, 1975.
[13] H. Renon and J. M. Prausnitz, "Local Compositions in Thermodynamic Excess
Functions for Liquid Mixtures," AIChE journal, vol. 14, pp. 135-144, 1968.
[14] C. C. Chen, H. Britt, J. Boston, and L. Evans, "Local Composition Model for
Excess Gibbs Energy of Electrolyte Systems. Part I: Single Solvent, Single
Completely Dissociated Electrolyte Systems," AIChE Journal, vol. 28, pp. 588-
596, 1982.

160
Texas Tech University, Nazir Hossain, August 2018

[15] C.-C. Chen and L. B. Evans, "A Local Composition Model for the Excess Gibbs
Energy of Aqueous Electrolyte Systems," AIChE Journal, vol. 32, pp. 444-454,
1986.
[16] A. Fredenslund, R. L. Jones, and J. M. Prausnitz, "Group‐Contribution Estimation
of Activity Coefficients in Nonideal Liquid Mixtures," AIChE Journal, vol. 21,
pp. 1086-1099, 1975.
[17] A. Klamt, "Conductor-Like Screening Model for Real Solvents: a New Approach
to the Quantitative Calculation of Solvation Phenomena," The Journal of Physical
Chemistry, vol. 99, pp. 2224-2235, 1995.
[18] S.-T. Lin and S. I. Sandler, "A Priori Phase Equilibrium Prediction from a
Segment Contribution Solvation Model," Industrial & Engineering Chemistry
Research, vol. 41, pp. 899-913, 2002.
[19] C.-C. Chen and Y. Song, "Solubility Modeling with a Nonrandom Two-Liquid
Segment Activity Coefficient Model," Industrial & Engineering Chemistry
Research, vol. 43, pp. 8354-8362, 2004.
[20] Y. Song and C.-C. Chen, "Symmetric Electrolyte Nonrandom Two-Liquid
Activity Coefficient Model," Industrial & Engineering Chemistry Research, vol.
48, pp. 7788-7797, 2009.
[21] N. Calvar, B. González, E. Gómez, and Á. Domínguez, "Vapor–Liquid Equilibria
for the Ternary System Ethanol + Water + 1-Ethyl-3-methylimidazolium
Ethylsulfate and the Corresponding Binary Systems Containing the Ionic Liquid
at 101.3 kPa," Journal of Chemical & Engineering Data, vol. 53, pp. 820-825,
2008.
[22] J. F. Wang, C. X. Li, and Z. H. Wang, "Measurement and Prediction of Vapor
Pressure of Binary and Ternary Systems Containing 1-Ethyl-3-
methylimidazolium Ethyl Sulfate," Journal of Chemical & Engineering Data, vol.
52, pp. 1307-1312, 2007.
[23] L. D. Simoni, L. E. Ficke, C. A. Lambert, M. A. Stadtherr, and J. F. Brennecke,
"Measurement and Prediction of Vapor−Liquid Equilibrium of Aqueous 1-Ethyl-
3-methylimidazolium-Based Ionic Liquid Systems," Industrial & Engineering
Chemistry Research, vol. 49, pp. 3893-3901, 2010.
[24] A. V. Orchillés, P. J. Miguel, E. Vercher, and A. Martínez-Andreu, "Isobaric
Vapor−Liquid Equilibria for 1-Propanol + Water + 1-Ethyl-3-methylimidazolium
Trifluoromethanesulfonate at 100 kPa," Journal of Chemical & Engineering
Data, vol. 53, pp. 2426-2431, 2008.
[25] S. K. Bhattacharia and C.-C. Chen, "Thermodynamic Modeling of KCl+H2O and
KCl+NaCl+H2O Systems Using Electrolyte NRTL Model," Fluid Phase
Equilibria.
[26] S. K. Bhattacharia, N. Hossain, and C.-C. Chen, "Thermodynamic Modeling of
Aqueous Na+–K+–Cl−–SO42− Quaternary System with Electrolyte NRTL Model,"
Fluid Phase Equilibria, vol. 403, pp. 1-9, 2015.
[27] S. K. Bhattacharia, S. Tanveer, N. Hossain, and C.-C. Chen, "Thermodynamic
modeling of aqueous Na+–K+–Mg2+–SO42− quaternary systems," Fluid Phase
Equilibria, vol. 403, pp. 1-9, 2015.

161
Texas Tech University, Nazir Hossain, August 2018

[28] A. plus, "Aspen Properties V8.4," ed. Burlington, MA, 2013: Aspen Technology,
Inc, 2013.
[29] P. Bonhôte, A.-P. Dias, N. Papageorgiou, K. Kalyanasundaram, and M. Grätzel,
"Hydrophobic, Highly Conductive Ambient-Temperature Molten Salts,"
Inorganic Chemistry, vol. 35, pp. 1168-1178, 1996.
[30] P. A. Hunt, "Why Does a Reduction in Hydrogen Bonding Lead to an Increase in
Viscosity for the 1-Butyl-2,3-dimethyl-imidazolium-Based Ionic Liquids?," The
Journal of Physical Chemistry B, vol. 111, pp. 4844-4853, 2007.
[31] K. Fumino, A. Wulf, and R. Ludwig, "Strong, Localized, and Directional
Hydrogen Bonds Fluidize Ionic Liquids," Angewandte Chemie International
Edition, vol. 47, pp. 8731-8734, 2008.
[32] Y. Zhang and E. J. Maginn, "The Effect of C2 Substitution on Melting Point and
Liquid Phase Dynamics of Imidazolium based-Ionic Liquids: Insights from
Molecular Dynamics Simulations," Physical Chemistry Chemical Physics, vol.
14, pp. 12157-12164, 2012.
[33] A. P. Abbott, "Model for the Conductivity of Ionic Liquids Based on an Infinite
Dilution of Holes," ChemPhysChem, vol. 6, pp. 2502-2505, 2005.
[34] C. Rey-Castro and L. F. Vega, "Transport Properties of the Ionic Liquid 1-ethyl-
3-methylimidazolium Chloride from Equilibrium Molecular Dynamics
Simulation. The Effect of Temperature," The Journal of Physical Chemistry B,
vol. 110, pp. 14426-14435, 2006.
[35] G. Yu, D. Zhao, L. Wen, S. Yang, and X. Chen, "Viscosity of Ionic Liquids:
Database, Observation, and Quantitative Structure‐Property Relationship
Analysis," AIChE Journal, vol. 58, pp. 2885-2899, 2012.
[36] H. Tokuda, S. Tsuzuki, M. A. B. H. Susan, K. Hayamizu, and M. Watanabe,
"How Ionic Are Room-Temperature Ionic Liquids? An Indicator of the
Physicochemical Properties," The Journal of Physical Chemistry B, vol. 110, pp.
19593-19600, 2006.
[37] D. R. MacFarlane, M. Forsyth, E. I. Izgorodina, A. P. Abbott, G. Annat, and K.
Fraser, "On the Concept of Ionicity in Ionic Liquids," Physical Chemistry
Chemical Physics, vol. 11, pp. 4962-4967, 2009.
[38] K. R. Harris, "Relations Between the Fractional Stokes−Einstein and
Nernst−Einstein Equations and Velocity Correlation Coefficients in Ionic Liquids
and Molten Salts," The Journal of Physical Chemistry B, vol. 114, pp. 9572-9577,
2010.
[39] W. Zhao, F. Leroy, B. Heggen, S. Zahn, B. Kirchner, S. Balasubramanian, et al.,
"Are There Stable Ion-Pairs in Room-Temperature Ionic Liquids? Molecular
Dynamics Simulations of 1-n-Butyl-3-methylimidazolium Hexafluorophosphate,"
Journal of the American Chemical Society, vol. 131, pp. 15825-15833, 2009.
[40] Y. Zhang and E. J. Maginn, "Direct Correlation between Ionic Liquid Transport
Properties and Ion Pair Lifetimes: A Molecular Dynamics Study," The Journal of
Physical Chemistry Letters, vol. 6, pp. 700-705, 2015.

162
Texas Tech University, Nazir Hossain, August 2018

[41] J. D. Tubbs and M. M. Hoffmann, "Ion-Pair Formation of the Ionic Liquid 1-


Ethyl-3-methylimidazolium bis(trifyl)imide in Low Dielectric Media," Journal of
Solution Chemistry, vol. 33, pp. 381-394, 2004.
[42] S. Schrodle, G. Annat, D. R. MacFarlane, M. Forsyth, R. Buchner, and G. Hefter,
"Broadband Dielectric Response of the Ionic Liquid N-methyl-N-
ethylpyrrolidinium Dicyanamide," Chemical Communications, pp. 1748-1750,
2006.
[43] M. G. Del Pópolo, C. L. Mullan, J. D. Holbrey, C. Hardacre, and P. Ballone, "Ion
Association in [bmim][PF6]/Naphthalene Mixtures: An Experimental and
Computational Study," Journal of the American Chemical Society, vol. 130, pp.
7032-7041, 2008.
[44] C. E. Bernardes, M. E. Minas da Piedade, and J. N. Canongia Lopes, "The
Structure of Aqueous Solutions of a Hydrophilic Ionic Liquid: the Full
Concentration range of 1-ethyl-3-methylimidazolium Ethylsulfate and Water,"
The Journal of Physical Chemistry B, vol. 115, pp. 2067-2074, 2011.
[45] J. Wang, W. Wang, P. A. Kollman, and D. A. Case, "Automatic atom type and
bond type perception in molecular mechanical calculations," Journal of molecular
graphics and modelling, vol. 25, pp. 247-260, 2006.
[46] J. Wang, R. M. Wolf, J. W. Caldwell, P. A. Kollman, and D. A. Case,
"Development and testing of a general amber force field," Journal of
computational chemistry, vol. 25, pp. 1157-1174, 2004.
[47] W. L. Jorgensen, J. Chandrasekhar, J. D. Madura, R. W. Impey, and M. L. Klein,
"Comparison of simple potential functions for simulating liquid water," The
Journal of chemical physics, vol. 79, pp. 926-935, 1983.
[48] D. J. Price and C. L. Brooks III, "A modified TIP3P water potential for simulation
with Ewald summation," The Journal of chemical physics, vol. 121, pp. 10096-
10103, 2004.
[49] M. P. Allen and D. J. Tildesley, Computer simulation of liquids: Oxford
university press, 1989.
[50] J.-P. Ryckaert, G. Ciccotti, and H. J. Berendsen, "Numerical integration of the
cartesian equations of motion of a system with constraints: molecular dynamics of
n-alkanes," Journal of Computational Physics, vol. 23, pp. 327-341, 1977.
[51] A. Jakalian, D. B. Jack, and C. I. Bayly, "Fast, efficient generation of high‐quality
atomic charges. AM1‐BCC model: II. Parameterization and validation," Journal
of computational chemistry, vol. 23, pp. 1623-1641, 2002.
[52] A. Jakalian, B. L. Bush, D. B. Jack, and C. I. Bayly, "Fast, efficient generation of
high‐quality atomic Charges. AM1‐BCC model: I. Method," Journal of
Computational Chemistry, vol. 21, pp. 132-146, 2000.
[53] Y. Zhang and E. J. Maginn, "A Simple AIMD Approach to Derive Atomic
Charges for Condensed Phase Simulation of Ionic Liquids," The Journal of
Physical Chemistry B, vol. 116, pp. 10036-10048, 2012.
[54] R. Hockney and J. Eastwood, "Computer Simulation Using Particles Taylor &
Francis," Inc., Bristol, UK, 1988.

163
Texas Tech University, Nazir Hossain, August 2018

[55] M. Parrinello and A. Rahman, "Polymorphic transitions in single crystals: A new


molecular dynamics method," Journal of Applied physics, vol. 52, pp. 7182-7190,
1981.
[56] W. Shinoda, M. Shiga, and M. Mikami, "Rapid Estimation of Elastic Constants
by Molecular Dynamics Simulation under Constant Stress," Physical Review B,
vol. 69, p. 134103, 2004.
[57] S. Plimpton, "Fast Parallel Algorithms for Short-Range Molecular Dynamics,"
Journal of Computational Physics, vol. 117, pp. 1-19, 1995.
[58] Y. Zhang, L. Xue, F. Khabaz, R. Doerfler, E. L. Quitevis, R. Khare, et al.,
"Molecular topology and local dynamics govern the viscosity of imidazolium-
based ionic liquids," The Journal of Physical Chemistry B, vol. 119, pp. 14934-
14944, 2015.
[59] E. Vercher, A. V. Orchillés, P. J. Miguel, and A. Martínez-Andreu, "Volumetric
and Ultrasonic Studies of 1-Ethyl-3-methylimidazolium
Trifluoromethanesulfonate Ionic Liquid with Methanol, Ethanol, 1-Propanol, and
Water at Several Temperatures," Journal of Chemical & Engineering Data, vol.
52, pp. 1468-1482, 2007.
[60] P.-H. Lin and R. Khare, "Local Chain Dynamics and Dynamic Heterogeneity in
Cross-Linked Epoxy in the Vicinity of Glass Transition," Macromolecules, vol.
43, pp. 6505-6510, 2010.
[61] S. Mani, F. Khabaz, R. V. Godbole, R. C. Hedden, and R. Khare, "Structure and
Hydrogen Bonding of Water in Polyacrylate Gels: Effects of Polymer
Hydrophilicity and Water Concentration," The Journal of Physical Chemistry B,
vol. 119, pp. 15381-15393, 2015.

164
Texas Tech University, Nazir Hossain, August 2018

CHAPTER 6
CONCLUSIONS AND FUTURE WORKS

We revisited the Pitzer model and eNRTL model for NaCl + H2O system to

investigate the concentration dependence of the model parameters at 298.15 K. The Pitzer

model and eNRTL model give significant divergence in the prediction of the mean ionic

activity coefficient at higher concentration of the electrolyte. The Pitzer model shows a

rapid increase in the mean ionic activity coefficient and eNRTL model reaches a plateau.

The molecular dynamics simulations and supersaturation data support the trend of the

eNRTL model. In future, other electrolyte systems will be investigated with molecular

dynamics simulations to predict the mean ionic activity coefficient and compare with the

model predictions, beyond the saturation limit.

The Pitzer model gives no hint of the temperature dependence rather uses totally

empirical nonlinear formulation and requires a very large number of adjustable

parameters. The eNRTL model requires a minimum number of binary interaction

parameters which are which can be expressed as Helmholtz type equation. We show that

the eNRTL model parameters can be correlated with excess Gibbs energy, enthalpy, and

heat capacity at 298.15 K and can be comparable with a comprehensive study.

We applied the eNRTL model for Li+–Na+–K+–Mg2+−SO42− quinary system as a


continuum effort to develop a comprehensive thermodynamic modeling for the

concentrated brine solutions. For this quinary system, the eNRTL model requires four

electrolyte-water and six electrolyte-electrolyte binary interaction parameters. The model

parameters, which involve Li2SO4, are regressed from the experimental osmotic

coefficient, excess enthalpy, and solubility data while other parameters are taken from the

literature. The eNRTL model represents various thermodynamic properties accurately

from 273.15–573.15 K and concentration up to saturation.

165
Texas Tech University, Nazir Hossain, August 2018

We extended the applicability of the eNRTL model to aqueous Ionic Liquids

(ILs). For thermodynamic modeling, Ionic liquids are often approximated as either in all

molecular form or completely dissociated in water. With conventional partial dissociation

chemistry, the thermodynamic models fail to predict the dissociation of ILs at infinite

dilution of ILs in water. We performed molecular dynamics simulations for three ionic

liquids (ILs): 1-ethyl-3methylimidazolium ethylsulfate, 1-ethyl-3-methylimidazolium

trifluoroacetate, and 1-ethyl-3-methylimidazolium trifluoromethanelsulfonate in water.

Consistent with the literature experimental data showing these ILs are partially

dissociated, the molecular dynamics simulations show the formation of singlets, doublets,

and triplets at very dilute concentration of ILs in water. Based on these findings, we

propose a dissociation chemistry for ILs and accurately correlate all available vapor

pressure and dissociation extent data over the entire IL concentration range for the three

ILs with symmetric electrolyte Non-Random Two Liquid model. In future, the clustering

mechanism of ILs in water will be investigated by density functional theory.

166

You might also like